Advances in Operations Research Volume 212, Article ID 94797, 32 pages doi:1.11/212/94797 Research Article A Hybrid Genetic Programming Method in Optimization and Forecasting: A Case Study of the Broadband Penetration in OECD Countries Konstantinos Salpasaranis and Vasilios Stylianakis Electrical and Computer Engineering Department, University of Patras, 26 Rio Patra, Greece Correspondence should be addressed to Konstantinos Salpasaranis, salpk@upatras.gr Received 23 March 212; Revised 21 June 212; Accepted 22 June 212 Academic Editor: Yi-Kuei Lin Copyright q 212 K. Salpasaranis and V. Stylianakis. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The introduction of a hybrid genetic programming method hgp in fitting and forecasting of the broadband penetration data is proposed. The hgp uses some well-known diffusion models, such as those of,, and, in the initial population of the solutions in order to accelerate the algorithm. The produced solutions models of the hgp are used in fitting and forecasting the adoption of broadband penetration. We investigate the fitting performance of the hgp, and we use the hgp to forecast the broadband penetration in OECD Organisation for Economic Co-operation and Development countries. The results of the optimized diffusion models are compared to those of the hgp-generated models. The comparison indicates that the hgp manages to generate solutions with high-performance statistical indicators. The hgp cooperates with the existing diffusion models, thus allowing multiple approaches to forecasting. The modified algorithm is implemented in the Python programming language, which is fast in execution time, compact, and user friendly. 1. Introduction Many methods have been proposed for predicting the penetration of new technology in a community. The subject has been described and analyzed by worldwide literature, extensively 1 6. Examples of the above methods are the diffusion models for the adoption of new technologies. The diffusion models are mathematical functions that follow an S-shaped curve in time. The diffusion models used in this study are the,, and 4.The parameters of the models have been estimated by regression analysis. Genetic algorithm GA is a probabilistic search method which uses the Darwinian principle of natural selection in finding an appropriate solution of a specific problem 7.
2 Advances in Operations Research GP is more general than GA, because the produced solution corresponds to a new program 8. The implementation of genetic programming GP in optimization problems has produced some important forecasting tools 7, 8. Generally, a GP begins with a set of initial randomly chosen functions solutions and this set is called population. A chromosome is a program solution of GP. Each solution has a fitness value, and this chromosome s fitness is evaluated. The next generation is the resultant of the Darwinian selection process. In this process, the best chromosomes, according to their fitness values, are selected for the next generation. Some of the selected chromosomes are randomly combined crossover and generate new chromosomes offspring. The mutation process also occurs, according to which a part of a randomly selected chromosome is changing. Finally, the chromosomes with better fitness values have better probability of being selected. The chromosomes of the new generation have better overall fitness value than those of the past generations. The whole process is repeated until an end condition has occurred 8, 9. This paper is structured as follows. At first, the diffusion models are shortly presented. The basic structure of the new modified GP hgp and its analysis follow. The next section contains the results analysis, and, finally, the conclusion is presented. In the Appendices A, B, andc, the syntax of some produced models and the statistic indicators as well as the estimation formulas of the modified GA are provided. 2. Diffusion Models Typically the diffusion process of innovative technologies in a society, such as broadband access, follows the sigmoid curve 6. In this paper, some well-known diffusion models have been used as follows. 2.1. Model The logistic diffusion model is the most known model among the sigmoid curves. The model is described by 2.1 : S y t, 2.1 1 a ef{t} where the diffusion of a new product in a society at time t is presented as y t. Also,f t a b t is a time-dependant function, and a, b are constant parameters. The S parameter is the limit of the function y t. When time t, then y t S. 2.2. Model The model was introduced as a law of mortality 1. The equation that we use is the known format of 2.2, as follows: y t S e ef t. 2.2
Advances in Operations Research 3 2.3 : We can also use the alternative format, with constant, which is described in y t S e ef t c, 2.3 where f t a b t is a time-dependent function and a, b, c are constant parameters. 2.3. Model The model proposes that the market for a new product consists of two major categories: innovators and imitators. Firstly, the innovators purchase the new product, and the imitators follow afterwards. This model s function is an extended logistic curve, where the cumulative adoption of the new technology y t for time t is presented in 2.4 : y t A C e B t 1 D e B t. 2.4 In 2.4, parameter A corresponds to initial purchasers of the new technology product. Parameter B is the sum of the innovators and imitators coefficients, p and q, respectively. This is presented as B p q. C parameter is C r p, where r is a constant and D parameter is D r q/a 4, 11. 3. Genetic Programming Method The specific hgp implements a hybrid strategy, which consists of two parts, the nonlinear regression analysis and the modified genetic algorithm part. The flowchart of Figure 1 shows the parts of the hgp. Firstly, a random set of solutions is created. Simultaneously, the candidate diffusion models are optimized by regression analysis and the produced models are inserted into the initial set of solutions. Now, the initial population of the hgp is prepared. Then, each solution is evaluated by its fitness function and the best solutions are inserted into a sorted list. The first, initial generation is ready. The termination criterion of the program is the maximum number of generations. After that, the selection of the best solutions takes place. It consists of the random combination of the best solutions, according to the crossover law and the random change of another randomly chosen solution. These solutions are going to be optimized by regression, and the next generation is ready. Finally, the whole process is repeated. 3.1. Solution Representation In GP, each chromosome is not a fixed-length character string but a program that is a possible solution to the problem 12. A chromosome in hgp, specifically, is presented as a string of characters in the Python programming language or as a parse tree. For example, the chromosome a b e c d t is presented in Figures 2 and 3 as string and parse tree, respectively 12.
4 Advances in Operations Research Random set of chromosomes Chromosomes of optimised diffusion models by regression Initial population Evaluation of population chromosomes NO Maximum generation? YES Results Selection Crossover Regression New population Mutation Figure 1: Flowchart of the hgp. Chromosome as a string of characters a + b E ( c d t ) Positions of the characters Figure 2: Representation of a chromosome in hgp as string.
Advances in Operations Research + a b E c d t Figure 3: Representation of a chromosome in hgp as a parse tree. The functions that are used in hgp are the addition, subtraction, protected division /, multiplication, and exponential E, because these are the same with the diffusion models functions. The parse tree consists of nodes. The terminal nodes are referred to as leaves. The leaves of the tree contain the variables or the constants of the program. As we can see the nonterminal nodes of the tree correspond to the function set of the hgp. 3.2. Initial Population The initial population is generated by the fusion of a randomly produced number of chromosomes and the diffusion models which are optimized by regression analysis. The initial diffusion models chromosomes are the optimized, family, and models. It should be noted that the diffusion models may differ according to the problem. The parameters of the initial diffusion models are optimized by nonlinear regression analysis, and the Levenberg-Marquardt algorithm has been used. From a programmer s point of view, each chromosome is a string data type which is presented as a parse tree in the internal structure of the Python programming language. The population appears as a list of strings. This list corresponds to the first generation for the program. 3.3. Evaluation During the evaluation process, we can use many functions as fitness indicators with real data. In the specific implementation, we use two different fitness functions as follows.
6 Advances in Operations Research Best value Sorted list according the fitness value of the selected chromosomes Fitness value 1st chromosome Fitness value 2nd chromosome Fitness value 3rd chromosome Fitness value 4th chromosome Worst value Fitness value Nth chromosome Figure 4: Representation of the sorted list of the selected chromosomes in hgp. In the fitting process, each chromosome is evaluated with the sum of squared error SSE, as in Appendix C in C.1. In forecasting, the evaluation function corresponds to the weighted sum of squared error wsse function, as in C.2. 3.4. Selection In hgp, the chromosomes with the best values of the fitness function smaller than a precision limit are inserted into a Python s list. The members of the list are sorted according to their fitness value. In Figure 4, we can see the structure of the sorted list. Then some randomly chosen chromosomes will be selected in order to produce the next generation using crossover and mutation processes. 3.. Crossover As it has been aforementioned, in the implementation of hgp, each chromosome is a string data type or a parse tree in the internal structure of the programming language. In the crossover process, two parents are randomly selected from the chromosomes list with the best fitness value. A crossover point is randomly chosen in the first parent string. Also another crossover point is randomly chosen in the second parent string. The first child offspring is generated when the substring of the first parent, which begins at the crossover point, is replaced by the substring from the second parent, which begins at the second crossover point. The second child is generated by the crossing over of the other parts substrings of the parents strings. In Figures and 6, the crossover operation is presented.
Advances in Operations Research 7 First parent a 1 + b 1 E ( c 1 d 1 t ) Crossover points Exchange substrings Second parent a 2 b 2 / E ( c 2 + d 2 t ) First child a 1 + ( c 2 + d 2 t ) Second child a 2 b 2 / E ( b 1 E ( c 1 d 1 t ) ) Figure : Crossover of the hybrid genetic programming method string representation. 3.6. Mutation In the mutation process, a chromosome is randomly chosen from the chromosomes list, with the best fitness value. The substring point, in which the mutation will take place, is also randomly chosen. The mutation replaces the function,, /,, E which is presented at the mutation point with a new random function in the substring. The mutation operation is presented in Figures 7 and 8. 4. Results In this section, the dataset, the statistic indicators, and the results of the study will be presented. The results will also be analysed in order to provide a satisfactory prediction for the broadband penetration in OECD countries. 4.1. Dataset In this study, the proposed method has been implemented on two different datasets. The first dataset presents the overall OECD broadband penetration. The data came from the OECD portal 13, which presents the total fixed wired broadband penetration in OECD countries. According to OECD reports, the overall broadband penetration is rapidly growing. The dataset concerns the time period from the second quarter Q2 of the year 22 until the forth quarter Q4 of the year 21. The dataset is comprised of 18 data points. The method is also implemented on the second dataset which concerns three innovators countries in fixed wired broadband technology adoption 14, namely, Sweden, The Netherlands, and Denmark. This dataset concerns the time period from the second quarter Q2 of the year 21 until the fourth quarter Q4 of 21 that is 2 data points.
8 Advances in Operations Research First parent Crossover point + a 1 b 1 c 1 E d1 t Second parent + a 2 b 2 E Crossover point c 2 d 2 t First child + Second child + a 1 a 2 c 2 b 2 E d 2 t b 1 E c 1 d 1 t Figure 6: Crossover of the hybrid genetic programming method parse tree representation. a + b E ( c d t ) Mutation point a b E ( c d t ) A randomly chosen function (*) has replaced the initial function (+) at the mutation point. Figure 7: Mutation of the hybrid genetic programming method string representation. + Mutation point a a b c E b A randomly chosen function ( ) has replaced the initial function (+) c E d t d t Figure 8: Mutation of the hybrid genetic programming method parse tree representation.
Advances in Operations Research 9 Table 1: Parameters initialization of hgp. Parameters of the hgp Maximum number of generations Evaluation function SSE Upper limit of the precision for the candidates for crossover and mutation. Table 2: Statistical indices of the first five produced hgp models OECD overall. Statistical indices of the hgp for OECD overall dataset Model name MAPE SSE MSE RMSE MAE hgp Model 1.7863 4.347E 2.1693E 6.186.131 hgp Model 2.9163 4.91669E 2.73149E 6.163.1271 hgp Model 3.1628 4.9931E 2.717E 6.166.1363 hgp Model 4.1628 4.9931E 2.717E 6.166.1363 hgp Model.1628 4.9932E 2.718E 6.166.1363 Thus, the fitting and forecasting ability of hgp are both tested in markets that have reached or are going to reach the saturation point of the penetration curve. 4.2. Statistic Indicators As mentioned above, the fitness function of each individual is the sum of squared error SSE for the fitting process. The results are analyzed by the estimation of some widely used statistical indices. The main indices of our analysis are the mean absolute percentage error MAPE, the mean square error MSE, the mean absolute error MAE, and the root mean square error RMSE. The statistic indicators of MAPE, MSE, MAE and RMSE are presented in Appendix B. 4.3. Fitting Results As aforementioned, the statistic indicator used for the fitting process is the SSE. In this section, the fitting performance of the generated models by the hgp as well as the comparison of the optimized diffusion models with the hgp s models is presented. Both hgp and optimized diffusion models concerning the fitting performance are presented in Appendix A.1. 4.3.1. Fitting Results for the Overall OECD Broadband Penetration Table 1 contains the initialization parameters for the execution of hgp concerning the whole dataset of the OECD countries. An example of the fitting performance for the first five hgp models, according to their fitness value SSE, is presented in Figure 9. The graphs represent the broadband penetration percentage y-axis and time x-axis as date. Each time point x-axis corresponds to a six-month period. The relative statistical indices SSE, MAPE, MSE, RMSE, and MAE of the produced hgp models are presented in Table 2. Also, the fitting performance of the optimized diffusion models, according to their fitness value SSE, is presented in Figure 1 and their statistical indices in Table 3.
1 Advances in Operations Research 2 hgp models-fitting performance (OECD) Broadband penetration (%) 2 1 1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Figure 9: The fitting performance of the first five hgp models OECD overall. 2 Diffusion models-fitting performance (OECD) Broadband penetration (%) 2 1 1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration (constant) Figure 1: The fitting performance of the optimized diffusion models for OECD overall dataset. Table 3: Representation of the statistics for the optimized diffusion models sorted by SSE. Statistical indices of the optimized diffusion models for OECD overall dataset Model name MAPE SSE MSE RMSE MAE.144.13461 7.47864E 6.273471.211296.144.134616 7.47867E 6.273472.211297 with constant.227.166627 9.277E 6.3424.2496.3391.316133 1.763E.41982.3731
Advances in Operations Research 11 Table 4: Parameters initialization of the hgp. Parameters of the hgp Maximum number of generations Evaluation function SSE Upper limit of the precision for the candidates for crossover and mutation. Table : Statistical indices of the first five produced hgp models Sweden. Statistical indices of the hgp for Sweden dataset Model name MAPE SSE MSE RMSE MAE hgp Model 1.867744 7.4676E 4.63861E 3.73378E 6.1932299 hgp Model 2.918124 8.9721E.39797E 4.4361E 6.21137 hgp Model 3.332876.26438 6.12299E 1.32179E.363641 hgp Model 4.484881.1968.1372 2.7984E.79212 hgp Model.486113.1746.1429 2.8728E.8632 Table 6: Representation of the statistics for the optimized diffusion models sorted by SSE Sweden. Statistical indices of the optimized diffusion models for Sweden dataset Model name MAPE SSE MSE RMSE MAE.827.1689 8.4E.9191.7991.827.1689 8.4E.9191.7991 with constant.83396.271.14.117.92.8642.31.178.13326.1173 When considering Tables 2 and 3, the hgp method achieves better statistical indices than those of the optimized diffusion models. For example, the first hgp model shows a SSE value of 4.347E and the model.13461. It should be noted that hgp was executed for generations, which is a rather medium number of generations. Thus, better results could be achieved for larger number of generations. 4.3.2. Fitting Results for Sweden, The Netherlands, and Denmark Broadband Penetration Table 4 contains the initialization parameters for the execution of hgp concerning the dataset for The Netherlands-Sweden-Denmark countries. The fitting performance for the first five hgp models, according to their fitness value SSE as well as the performance of the optimized diffusion models is presented in Figures 11 and 12 for Sweden, Figures 13 and 14 for The Netherlands, and Figures 1 and 16 for Denmark, respectively. The relative statistical indices SSE, MAPE, MSE, RMSE, and MAE of the produced hgp models are presented in Tables and 6 for Sweden, Tables 7 and 8 for The Netherlands, and Tables 9 and 1 for Denmark.
12 Advances in Operations Research 3 hgp models-fitting performance (Sweden) Broadband penetration (%) 3 2 2 1 1 Jun-1 Dec-1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Figure 11: The fitting performance of the first five hgp models for Sweden. 3 Diffusion models-fitting performance (Sweden) Broadband penetration (%) 3 2 2 1 1 Jun-1 Dec-1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration (constant) Figure 12: The fitting performance of the optimized diffusion models for Sweden. Table 7: Statistical indices of the first five produced hgp models The Netherlands. Statistical indices of hgp for The Netherlands dataset Model name MAPE SSE MSE RMSE MAE hgp Model 1.13729.18787 9.39E 6.36.2343 hgp Model 2.12842.18848 9.42E 6.37.234 hgp Model 3.1377.19367 9.68E 6.3112.2264 hgp Model 4.1377.19367 9.68E 6.3112.2264 hgp Model.198.2294 1.1E.318.21
Advances in Operations Research 13 4 hgp models-fitting performance (The Netherlands) Broadband penetration (%) 3 3 2 2 1 1 Jun-1 Dec-1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Figure 13: The fitting performance of the first five hgp models for The Netherlands. Broadband penetration (%) 4 3 3 2 2 1 1 Jun-1 Dec-1 Diffusion models-fitting performance (The Netherlands) Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration (constant) Figure 14: The fitting performance of the optimized diffusion models for The Netherlands. Table 8: Statistics for the optimized diffusion models sorted by SSE The Netherlands. Statistical indices of the optimized diffusion models for The Netherlands dataset Model name MAPE SSE MSE RMSE MAE.1446.23692 1.18E.3442.248.1447.23692 1.18E.3442.248 with constant.313.41264 2.6E.442.3828.7946.1367898 6.84E.827.71
14 Advances in Operations Research 4 hgp models-fitting performance (Denmark) Broadband penetration (%) 3 3 2 2 1 1 Jun-1 Dec-1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Figure 1: The fitting performance of the first five hgp models for Denmark. 4 Diffusion models-fitting performance (Denmark) Broadband penetration (%) 3 3 2 2 1 1 Jun-1 Dec-1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration (constant) Figure 16: The fitting performance of the optimized diffusion models for Denmark. According to Tables and 6, the hgp method achieves better statistical indices than those of the optimized diffusion models. For example, the first hgp model achieves an SSE value 7.4676E while the model an SSE value of.1689. Once again, according to Tables 7 and 8, the hgp method achieves better statistical indices than those of the optimized diffusion models. Finally, according to Tables 9 and 1, the hgp method achieves better statistical indices than those of the optimized diffusion models. The first hgp model achieves an SSE value of.66649 while the model.11171.
Advances in Operations Research 1 Table 9: Statistical indices of the first five produced hgp models Denmark. Statistical indices of hgp for Denmark dataset Model name MAPE SSE MSE RMSE MAE hgp Model 1.412.66649 2.83324E.323.4332 hgp Model 2.47466.99483 2.99742E.47.442 hgp Model 3.47487.99823 2.99912E.476.443 hgp Model 4.47489.9987 2.99938E.477.444 hgp Model.4731.672 3.286E.48.446 Table 1: Statistics for the optimized diffusion models sorted by SSE Denmark. Statistical indices of the optimized diffusion models for Denmark dataset Model name MAPE SSE MSE RMSE MAE.49134.11171.E.742.6494.49134.11174.E.742.6494 with constant.49134.11171.e.742.6494.49134.11174.e.742.6494 Table 11: Initialization parameters of hgp. Parameters of the hgp Maximum number of generations Evaluation function wsse Upper limit of the precision for the candidates for crossover and mutation.9 4.4. Forecasting Results The forecasting results of the generated models by hgp are presented in this section as well as the comparison of the optimized diffusion models with the hgp s models. The statistic indicator used for the forecasting process is the wsse. Both hgp and the optimized diffusion models concerning the forecasting performance are presented in Appendix A.2. 4.4.1. Forecasting Results for the Overall OECD Broadband Penetration Table 11 contains the initialization parameters for the execution of hgp for the forecasting process. The whole dataset of the OECD countries contains 18 data points 2 points per year since 22 until 21. The forecasting method for a 2-year prediction uses 14 points data for hgp training of the dataset and implements the statistical indices to this 14-point subset. The graph of the forecasting performance is shown in Figure 17. In every graph, the forecast period window is presented into the blue rectangle.
16 Advances in Operations Research Table 12: Statistical indices of the first five produced forecasting hgp models for OECD overall. Statistical indices of the hgp for OECD forecasting 14 points training Model name MAPE wsse MSE RMSE MAE hgp Model 1.9744 3.6791E 4.2E 6.2.1412 hgp Model 2.14944 4.2E 4.82E 6.219.168 hgp Model 3.1427 4.281E.38E 6.2319.1683 hgp Model 4.19881.239E 6.6E 6.279.241 hgp Model.26.4818E 6.86E 6.262.263 2 hgp models-forecasting performance (OECD) Broadband penetration (%) 2 1 1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Figure 17: The forecasting performance of hgp forecast period 2 years ahead. The relative statistical indices, for the training points, of the produced forecasting hgp models are presented in Table 12. The forecasting performance of the optimized diffusion models, according to their fitness value wsse for the 14 training points, is presented in Figure 18 and their statistical indices in Table 13. Considering Tables 12 and 13, the conclusion is that the hgp method achieves better statistical indices than those of the optimized diffusion models. We can see that the first hgp model achieves a wsse value 3.6791E while the model 6.1366E. The comparison of the models residuals against time data points, especially for the 4 last data points the forecast period, shows the predominance of the hgp models see Figure 19. 4.4.2. Forecasting Results for the Sweden-The Netherlands-Denmark Broadband Penetration Table 14 contains the initialization parameters for the execution of hgp in the forecasting process. The datasets of Sweden, The Netherlands, and Denmark contain 2 data points 2 points per year since 21 until 21. The forecasting method for a 2-year prediction uses 16 data points for hgp training of the dataset and implements the statistical indices to the subset
Advances in Operations Research 17 3 Diffusion models-forecasting performance (OECD) Broadband penetration (%) 2 2 1 1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration (constant) Figure 18: The forecasting performance of the optimized diffusion models forecast period 2 years ahead..8 Forecasting performance-statistical indices (OECD).7.6..4.3.2.1 hgp Model 1 hgp Model hgp Model 2 3 hgp Model hgp Model 4 (constant) MAPE.9977.9977.8944.8944.116397.233423.7421.4198287.23319 MAE.218322.218322.291.291.27637.48388.1767467.99694.48412 Figure 19: The forecasting performance of the models forecast period 2 years ahead. of the training points. The graphs of the forecasting performance of the hgp and optimized diffusion models are shown in Figures 2 and 21 for Sweden, Figures 23 and 24 for The Netherlands, and Figures 26 and 27 for Denmark, respectively. The relative statistical indices for the training data are presented in Tables 1 and 16 for Sweden, Tables 17 and 18 for The Netherlands, and Tables 19 and 2 for Denmark. Finally, the statistical indices, MAPE and MAE, that describe the forecasting performance of the models in a forecasting horizon of 2-years are presented in Figures 22, 2,and28 for the three countries.
18 Advances in Operations Research Table 13: Representation of the statistics for the optimized diffusion models sorted by wsse forecast period 2 years ahead. Statistical indices of the optimized diffusion models for OECD forecasting 14-point training Model name MAPE wsse MSE RMSE MAE.16716 6.1366E 7.18E 6.2679.22.16716 6.1373E 7.18E 6.2679.22 with constant.211 6.9844E 8.1E 6.2917.2349.22379 9.999E 1.8E.3284.2678 Table 14: Parameters initialization of hgp. Parameters of the hgp Maximum number of generations Evaluation function wsse Upper limit of the precision for the candidates for crossover and mutation.9 3 hgp models-forecasting performance (Sweden) Broadband penetration (%) 3 2 2 1 1 Jun-1 Dec-1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Figure 2: The forecasting performance of the first five hgp models for Sweden forecast period 2 years ahead. (1) Forecasting Results for Sweden Broadband Penetration Considering Tables 1 and 16, it is concluded that the hgp method again achieves better statistical indices than those of the optimized diffusion models for the training subset. The first hgp model has a wsse value of.3e while the model.433. It should be mentioned that the hgp achieves a satisfactory performance after the 16th data point, with minimized residuals errors, MAPE and MAE for the last 4 data points 2-year forecast horizon, see Figure 22. (2) Forecasting Results for The Netherlands Broadband Penetration Further commenting on the forecasting results, according to Tables 17 and 18, it is concluded that the hgp method again achieves better statistical indices than those of the optimized
Advances in Operations Research 19 4 Diffusion models forecasting performance (Sweden) Broadband penetration (%) 3 3 2 2 1 1 Jun-1 Dec-1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration (constant) Figure 21: The forecasting performance of the optimized diffusion models for Sweden forecast period 2 years ahead..18 Forecasting performance-statistical indices (Sweden).16.14.12.1.8.6.4.2 hgp Model 1 hgp Model 2 hgp Model 3 hgp Model hgp Model 4 (constant) MAPE.169673.16729.1748692.1672911.171729.16481.1672213.1212988.164641 MAE.24831.2468.2892.293.42911.317676.2816.382664.31769 Figure 22: The forecasting performance of the models for Sweden forecast period 2 years ahead. diffusion models for the 16-point training subset. The first hgp model shows a wsse value of.1 while the and models 7.31E-. Both parts, the hgp and diffusion models, specially and models, achieve well enough performance after the 16th data point, with minimized residuals errors for the last 4 data points 2-year forecast horizon, see Figure 2.
2 Advances in Operations Research 4 hgp models-forecasting performance (The Netherlands) Broadband penetration (%) 3 3 2 2 1 1 Jun-1 Dec-1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Figure 23: The forecasting performance of the first five hgp models and the optimized diffusion models for The Netherlands forecast period 2 years ahead. Broadband penetration (%) Diffusion models-forecasting performance (The Netherlands) 4 4 3 3 2 2 1 1 Jun-1 Dec-1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration (constant) Figure 24: The forecasting performance of the optimized diffusion models for The Netherlands forecast period 2 years ahead. Table 1: Statistical indices of the first five produced hgp models Sweden forecast period 2 years. Statistical indices of hgp for Sweden dataset 16-point training Model name MAPE wsse MSE RMSE MAE hgp Model 1.1114.3E 4.81E 6.2192.1692 hgp Model 2.893.66E 4.71E 6.2171.146 hgp Model 3.84.67E 4.69E 6.2166.138 hgp Model 4.81.69E 4.74E 6.2176.149 hgp Model.811.79E 4.81E 6.2192.17
Advances in Operations Research 21.4 Forecasting performance-statistical indices (The Netherlands).4.3.3.2.2.1.1. hgp Model 1 hgp Model 2 hgp Model hgp Model hgp Model 3 4 (constant) MAPE.97734.9694.969449.96922.96949.1491476.427693.1718236.1491469 MAE.3663.3688.3688.3687.36882.674.16924.646962.671 Figure 2: The forecasting performance of the models for The Netherlands forecast period 2 years ahead. 4 hgp models-forecasting performance (Denmark) Broadband penetration (%) 3 3 2 2 1 1 Jun-1 Dec-1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Figure 26: The forecasting performance of the first five hgp models for Denmark forecast period 2 years ahead. (3) Forecasting Results for Denmark Broadband Penetration Finally, according to Tables 19 and 2, once again the hgp method achieves better statistical indices than those of the optimized diffusion models for the training subset. The first hgp
22 Advances in Operations Research Broadband penetration (%) Diffusion models-forecasting performance (Denmark) 4 4 3 3 2 2 1 1 Jun-1 Dec-1 Jun-2 Dec-2 Jun-3 Dec-3 Jun-4 Dec-4 Jun- Dec- Jun-6 Dec-6 Jun-7 Dec-7 Jun-8 Dec-8 Jun-9 Dec-9 Jun-1 Dec-1 Broadband penetration (constant) Figure 27: The forecasting performance of the optimized diffusion models for Denmark forecast period 2 years ahead..12 Forecasting performance-statistical indices (Denmark).1.8.6.4.2 hgp Model 1 hgp Model 2 hgp Model 3 hgp Model hgp Model 4 (constant) MAPE.719464.116.8613.86899.86899.279669.112494.941679.61841 MAE.2694.393978.323139.322913.322912.196271.411334.3338.2299348 Figure 28: The forecasting performance of the optimized diffusion models for Denmark forecast period 2 years ahead. model achieves a wsse value of 3.396E while the model.319962. The hgp achieves well enough performance after the 16th data point, with minimized residuals errors, MAPE and MAE for the last 4 data points 2-year forecast horizon, see Figure 28.
Advances in Operations Research 23 Table 16: Statistics for the optimized diffusion models sorted by wsse Sweden forecast period 2 years. Statistical indices of the optimized diffusion models for Sweden dataset 16-point training Model name MAPE wsse MSE RMSE MAE.439.433.3E.789.6376.439.433.3E.789.6376 with constant.6964.13 6.8E.8114.7267.488.82 8.47E.922.7889 Table 17: Statistical indices of the first five produced hgp models for The Netherlands forecast period 2 years. Statistical indices of hgp for The Netherlands dataset 16-point training Model name MAPE wsse MSE RMSE MAE hgp Model 1.14673.1 8.8E 6.2929.2194 hgp Model 2.1729.13 8.84E 6.2973.227 hgp Model 3.1729.13 8.84E 6.2973.227 hgp Model 4.17223.13 8.8E 6.297.2271 hgp Model.1723.13 8.86E 6.2976.2272 Table 18: Statistics for the optimized diffusion models sorted by wsse for The Netherlands forecast period 2 years. Statistical indices of the optimized diffusion models for The Netherlands dataset 16-point training Model name MAPE wsse MSE RMSE MAE.18228 7.31E 7.36E 6.2712.269.18228 7.31E 7.36E 6.2712.269 with constant.3332.162 1.76E.4199.32.7421.437.84E.7639.6336. Conclusions This paper introduces a new GP method that produces fitting and forecasting solutions models with well enough performance. The whole process is assisted by the insertion of some widely used diffusion models like, family, and, in the initial population of the chromosomes. The hgp method was implemented with a dataset concerning the overall broadband penetration of the OECD countries, as well as the datasets of Sweden, The Netherlands, and Denmark which are pioneers in broadband technologies. Both the fitting and the forecasting performance of the method presented satisfactory statistical indices. The proposed method differs from the classic GP method in several points. First, some well-known diffusion models,, family, and, which have been optimized by regression analysis, have been inserted in the randomly generated initial population. Second, the regression process has been implemented for each chromosome and after each crossover and mutation operation in order to maximize the algorithm s efficiency. Also, the crossover and mutation processes are implemented by a random
24 Advances in Operations Research Table 19: Statistical indices of the first five produced hgp models for Denmark forecast period 2 years. Statistical indices of the hgp for Denmark dataset 16-point training Model name MAPE wsse MSE RMSE MAE hgp Model 1.24844 3.396E 7.2E 6.2692.218 hgp Model 2.2488 6.2484E 1.2E.321.2468 hgp Model 3.2291 6.421E 1.E.3233.2483 hgp Model 4.22944 6.42636E 1.E.3233.2483 hgp Model.22944 6.42636E 1.E.3233.2483 Table 2: Statistics for the optimized diffusion models sorted by wsse for Denmark forecast period 2 years. Statistical indices of the optimized diffusion models for Denmark dataset 16-point training Model name MAPE wsse MSE RMSE MAE.897.319962 4.33E.682.79.4866.36221 4.14E.6434.734 with constant.776.62689 6.2E.776.77.429.683336 6.6E.812.6943 selection of individuals from a sorted list which contains the chromosomes with the best performance. In general, the produced hgp could be considered as a tool which generates solutions with well enough performance in fitting and forecasting of the broadband penetration. In this paper, the forecasting horizon of hgp was two years ahead. A further investigation of the hgp performance for a longer forecast horizon would be desirable. It should be noted that the hgp method performance could be further improved with the insertion of more functions into the function set. A future study with an enrichment of the function set of the hgp could thus be considered. Appendices A. In this section, some of the produced hgp models as well as the optimized diffusion models concerning the fitting and the forecasting performance that were analysed before are presented. Each produced model corresponds to a Python s string data type of the program. A.1. Fitting Performance Models For more details, see Tables 21, 22, 23, 24, 2, 26, 27,and28. A.2. Forecasting Performance Models For more details, see Tables 29, 3, 31, 32, 33, 34, 3,and36.
Advances in Operations Research 2 Table 21: Representation of the first five produced hgp models in fitting OECD overall. Model name hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Fitting performance hgp models for OECD overall Model.3711429972 E 3.186992821 E.17931441739 t E E 4.31337138346e 11 E.394322724 t 3.81888339 t.292663146 3.813611136 t.314714673 E 3.1134946 E.2472629962 t E.168943728131/ 1 E.623181732.7336313276 t.29797486322 E 3.4327778 E.2327711194 t E.12721433893 E 1981.43188/ 1 E 4.482143374.41892198 t.297974843236 E 3.4327 E.23277113 t E.12721423417 E 7692.989479/ 1 E 4.429746.41896114 t.29797468462 E 1.837714382 E.2327721971 t E 1.22871767 E.232769748624 t E.12721281 E 22946.64436/ 1 E 38.914434992.41889939 t Table 22: Representation of the optimized diffusion models in fitting OECD overall. Fitting performance optimized diffusion models for OECD overall Model name with constant Model.26746286667/ 1 E 2.833876374.419392431 t.26746298117 1 E 4.369337298e 8.41939129613 t 267.1843419 / 1.41939129613/4.369337298e 8 E 4.369337298e 8.41939129613 t 267.1843419.24779967624 E 6.93667714 E.312988127799 t.323712682778.31117446387 E 3.4314971411 E.2318132619 t Table 23: Representation of the first five produced hgp models in fitting Sweden. Fitting performance hgp models for Sweden Model name hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Model.316199143/ 1 E 24.361337631.248179188862 t 286.8627612.733667168 7632.2371 264.698228 t 188.226187141 1/ 1 E 11.21949426.269181697779 t.6323289913/ 1 E 89.981469 13.668273689 t / 1 E.2468629376.1127817324 t 64.28699231864.791617138 1.872383928 t.32228377413/ 1 E.238624687.297883897 12.8929141638 134.98971326 2.793439732 98.6391827.34439312981 t / 1 E.312431182.4611246618.1186199978 t.938838216.317318163/ 1 E 4.86189319 1.46227222 2.17969243374 E 22.362926623/ 1 E.7911748628.8739477331 t.3173296432 1/ 1 E 6.287814917.23176477 E 22.2379974 21.877663 42.62973363/ 1 E 3.896242943.288888424.9916377 t
26 Advances in Operations Research Table 24: Representation of the optimized diffusion models in fitting Sweden. Fitting performance optimized diffusion models for Sweden Model name Model.33628292699/ 1 E 2.6246214.1847273171 t.33628296867 1 E 2.433762338e 8.1847467378 t 231.6483773 / 1.1847467378/2.433762338e 8 E 2.433762338e 8.1847467378 t 231.6483773 with.2889126679 constant E 8.2983987 E.44628448837 t.79684.36638197367 E 3.314128436 E.314112794 t Table 2: Representation of the first five produced hgp models in fitting The Netherlands. Model name hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Model Fitting performance hgp models for The Netherlands.3282974423 E 1.628344 1.1126326886 E.3899834891 E.6812496293 t E 32.61777674 E 2.6392948.444872879 t E E.38212766 t E E 6.163667129 E 7.7643494 E 1.862862714 19.91619767 E 2.1924988 E.441172397 t 1.133883482.272274693312 E 1.48798793 4.3977447776 E 1.487793796e 6 E.71248718888 t E 2.1748492 E 2.1499998.376142836 t E E.18893694928 t E E 9.713187786 E 697.169312 12.774228698 E.127948362 t E.88213942 t 2.188443226.2799682383 E 1.87788178.69167161119 E.937233488237 E.147416224 t E 29.83474333 E 2.6237934.176934827 t E E.36681637 t E E 2.7938314693 E 14.667728363 E 3.4181431 64.4933792 E 3388.8924389 E.17411874631 t 31.29726778.2799682383 E 1.87788178.69167161119 E.937233488237 E.147416224 t E 29.83474333 E 2.6237934.176934827 t E E.36681637 t E E 2.7938314693 E 14.667728363 E 3.4181431 64.4933792 E 3388.8924389 E.17411874631 t 31.29726778.6229772664 E E 67628.39866 t E E.98693387 E.24697777946 t E.1113764774 t E E.8642471227 t E E 1844.32674 E.147819974 t E.39844163448 t E E 2.7368114239 E t E 3.3989316 E.4717412 t 1.12649641 B. The MAPE is presented in B.1. The sum is over time period t 1,...,T, r t is the raw data actual value for time t, andy t is the estimated model s value: MAPE 1 T T r t y t r t. t 1 B.1
Advances in Operations Research 27 Table 26: Representation of the optimized diffusion models in fitting The Netherlands. Fitting performance optimized diffusion models for The Netherlands Model name with constant Model.383817447797/ 1 E 2.99687319.616971349 t.383817124 1 E 4.19618494e 8.616836913 t 18.776347814 / 1.616836913/4.19618494e 8 E 4.19618494e 8.616836913 t 18.776347814.362686732 E 7.28281633264 E.473878131 t.3218647123.4863171889 E 4.42176988 E.3731646688 t Table 27: Representation of the first five produced hgp models in fitting Denmark. Fitting performance hgp models for Denmark Model name hgp Model 1 Model.63426994728 E 1618.49488 E 1.292199213 E 3.43879264146 E 1.2876774838 E E 3934393.2921 E 19.8648947917 E.143233298 E.7439262374 E E.434184249.1849278434.1424687633 t hgp Model 2 1.476267797 E 688787.219 E 17.6216887 E.32894643732 E.93778168 E E.361321.249644863.126941999487 t hgp Model 3 1.49461946 E 221143.91 E 16.689296686 E.34886473888 E.93762881193 E E.7243332.77878769.12669874873 t hgp Model 4 1.1783993 E 2197484.4721 E 16.136349 E.321627827 E.938394729 E E.2938781436.223636288.1266922938 t hgp Model 1.62443316 E 43372.9296 E 14.91611718 E.3941661739 E.94419376422 E E.2712913.28969466.487487926.1264614486 t Table 28: Representation of the optimized diffusion models in fitting Denmark. Fitting performance optimized diffusion models for Denmark Model name with constant Model.38671712691/ 1 E 2.76922996844.8481412812 t.3867182467 1 E 7.4833939618e 8.8481196298 t 184.676718196 / 1.8481196298/7.4833939618e 8 E 7.4833939618e 8.8481196298 t 184.676718196.3641894712 E 6.8132991138 E.42818612 t.3392427247.4111161742 E 3.8644149696 E.364379923 t
28 Advances in Operations Research Table 29: Representation of the first five produced hgp models in forecasting OECD overall. Forecasting performance hgp models for OECD overall Model name hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 Model.26669166828 1 E 4.8998293 E.3918843678 E 6.764212976e 7 t E.36444388831 t E 32.92944994 E 14.939483737 E.1486124973 t t 2.1667676882.26669166828 1 E 4.8998293 E.3918843678 E 6.764212976e 7 t E.36444388831 t E 32.92944994 E 14.939483737 E.1486124973 t t 2.1667676882.264888163 1 E 2.74789639 E 1.4363179622 E.234172233 t E.4616437 t E.6831376874 E.498481378 E 6.36227697e 6 t E.6831376874 E.498481378.12362874712 t 2243.8982316.264888163 1 E 2.74789639 E 1.4363179622 E.234172233 t E.4616437 t E.6831376874 E.498481378 E 6.36227697e 6 t E.6831376874 E.498481378.12362874712 t 2243.8982316 hgp Model.38668146/ 1 E 4.931261976 E 297363.4221 E 13.331328294 E.1471318 t E.61488398 E 1 16.2618967 t 284.9984 Table 3: Representation of the optimized diffusion models in forecasting OECD overall. Forecasting performance optimized diffusion models for OECD overall Model name Model.27791816466/ 1 E 2.869789463.412172 t.2779119992 1 E 1.4318179e 7.4128748 t 24.79644296 / 1.4128748/1.4318179e 7 E 1.4318179e 7.4128748 t 24.79644296 with.2914197222 constant E 4.92811622 E.29238386 t.2381918223.3994721399 E 3.32916469 E.1827834374 t MSE, MAE, and RMSE are presented in B.2, B.3, and B.4, respectively: [ ] T 2 r t y t MSE, B.2 T t 1 MAE 1 T r t y t, B.3 T t 1 [ ] RMSE T 2 r t y t MSE. T t 1 B.4
Advances in Operations Research 29 Table 31: Representation of the first five produced hgp models in forecasting Sweden. Forecasting performance hgp models for Sweden Model name hgp Model 1 Model 7.292738763/ 1 E 43.99964917 E 2.72936112 2.8349899 8.18376223 E 1.79173227 E 1.7472448 1.218976128.46422497147 E 1.2239739844 E.1413947816 t E.7493827 E.247828367 t hgp Model 2.31128113 E 8.2716729973 E 1.683189 E.3272789748 E.1196936679 t.61418876 1774.1877418 E 4.941741619 E E.4329282 t hgp Model 3.3972977447 E 3.32613666 E 1828787.414 E 21.39826 E.296222489 t.38778147.27286148 t E E E 21239.3614 E 13.33181431 E.3878493444 t hgp Model 4.3996868 E 9.8732389121 E 1.619468498 E.76117793 E.118396422714 t.3628776298 2244.3177439 E 4.7968896186 E E.396787831 t 3.141788827e hgp Model.39826417 E 14.6146178732 E 1.72422423849 E.2189997771 E.1168633386 t.4378829643 92.4849 E.96338688 E E.293446778 t Table 32: Representation of the optimized diffusion models in forecasting Sweden. Forecasting performance optimized diffusion models for Sweden Model name Model.373474846/ 1 E 2.124733322.4466416 t.373481719 1 E 4.161391e 8.4464711241 t 21.43997821 / 1.4464711241/4.161391e 8 E 4.161391e 8.4464711241 t 21.43997821 with constant.3714868271 E.39879234 E.31347376267 t.37384813167.493817739 E 3.13866616 E.21471247 t C. In the fitting process, each chromosome is evaluated with the sum of squared error SSE, as in T [ ] 2. SSE r t y t C.1 t 1
3 Advances in Operations Research Table 33: Representation of the first five produced hgp models in forecasting The Netherlands. Forecasting performance hgp models for The Netherlands Model name hgp Model 1 Model.23326242314 E 2.488891372 E.899289313 t E 23.1279794 E.6489882 t.21773392 hgp Model 2.383942499 1 E 1.73831772618e 8.6173396621 t 19.6814628 / 1.6173396621/1.73831772618e 8 E 1.73831772618e 8.6173396621 t 19.6814628 hgp Model 3 hgp Model 4.383948724/ 1 E 3.2678413.617339638782 t.3836192777/ 1 E 3.33671841 3.313937846 E 6.9744362173e 6 t E.67696636631 hgp Model 6.998328774e / 1 1.1824284 E.36789483 E.617182268173 t.144691744 t E 642.7237678.74984366 t Table 34: Representation of the optimized diffusion models in forecasting The Netherlands. Forecasting performance optimized diffusion models for The Netherlands Model name Model.378982831/ 1 E 3.1882999246.627143622 t.378983331 1 E 2.34739324e 8.6271444871 t 187.8294969 / 1.6271444871/2.34739324e 8 E 2.34739324e 8.6271444871 t 187.8294969 with constant.36821712377 E 6.91399399 E.4423117 t.3818877983.433117173 E 4.1487467224 E.3427136728 t In C.1, the sum is over the time period t 1,...,T.Also,r t is the real data for time t,andy t is the model s value 1. In forecasting, the evaluation function corresponds to the weighted sum of squared error wsse function, as in T [ ] 2. wsse w t r t y t t 1 C.2 In this function, a weight w t t/t is used, in order to focus on the time interval near the last observed or training data.
Advances in Operations Research 31 Table 3: Representation of the first five produced hgp models in forecasting Denmark. Forecasting performance hgp models for Denmark Model name hgp Model 1 hgp Model 2 hgp Model 3 hgp Model 4 hgp Model Model.37274249473/ 1 E 1.4932887728 1.6616878321.6781686919.6782919719 t t 4744.3143422 26.3429468 t 46.128987279 t 2.819328343 47.31389938 46.128987279 t/ 1 2.9489742172/ 14.4928944 E E E E 69.3366648 1.18868672 t.367676617/ 1 E 1.4326692 1.681441948.6768786124.6766477973 t t 6772.6811399 7863.139296 7863.17392 23.241389 t 9.11367676744.6774318948.6783923 t t 9622.19784 3.71296243976 t 13.38146941 t 69.433186343.3687922848/ 1 E 1.916211334 1.737867616.6768461973.67491287 t t 9961.43677147.63213847.6492811262 t t.4178111612 t 8794.16681 16119.989637 t 11.39792771.3687966886/ 1 E 1.632186 1.77333273.64733227916.613174937 t t.99762183839 t 1381.881184 111.3767878 1189.838866.661623277964.667282974 t t 467.17819 3972.8191893 t 118.6282811.3687966931/ 1 E 1.926199791 1.737217272.64378132468.62611789 t t 726.87441262 74.12297264.6617443967.64191722 t t 148193.4147 61.17627868 624.34691 t/ 1 4.141341813/.6237184782 E E 26.4687228.12122244894 t Table 36: Representation of the optimized diffusion models in forecasting Denmark. Forecasting performance optimized diffusion models for Denmark Model name with constant Model.472229441/ 1 E 2.789787126.4636884 t.41272936689 1 E.14382912913.49432328 t 2.438191217 / 1.49432328/.14382912913 E.14382912913.49432328 t 2.438191217.43391123761 E 4.3469849161 E.337136649 t.1997994121.4762814231 E 3.31281642 E.291742978 t Acknowledgment The authors wish to express their acknowledgments to Professor Yi-Kuei Lin, National Taiwan University of Science and Technology, Taiwan, for his constructive comments and suggestions, which helped to improve the quality of this paper.
32 Advances in Operations Research References 1 N. Meade and T. Islam, Modelling and forecasting the diffusion of innovation a 2-year review, International Journal of Forecasting, vol. 22, no. 3, pp. 19 4, 26. 2 Z. Griliches, Hybrid corn: an exploration in the economics of technological change, Econometrica, vol. 2, no. 4, pp. 1 22, 197. 3 E. Mansfield, Technical change and the rate of imitation, Econometrica, vol. 29, pp. 741 766, 1961. 4 F. M., A new product growth for model consumer durables, Management Science, vol., no. 12, pp. 182 1832, 24. S. Konstantinos and S. Vasilios, A new empirical model for short-term forecasting of the broadband penetration: a short research in Greece, Modelling and Simulation in Engineering, vol. 211, Article ID 79896, 1 pages, 211. 6 E. M. Rogers, Diffusion of Innovations, The Free Press, New York, NY, USA, th edition, 23. 7 J. H. Holland, Adaptation in Natural and Artificial Systems, University of Michigan Press, Ann Arbor, Mich, USA, 197. 8 J. R. Koza, Genetic programming as a means for programming computers by natural selection, Statistics and Computing, vol. 4, no. 2, pp. 87 112, 1994. 9 W. Lee and H. Y. Kim, Genetic algorithm implementation in Python, in Proceedings of the 4th Annual ACIS International Conference on Computer and Information Science (ICIS ), pp. 8 12, July 26. 1 C. Christodoulos, C. Michalakelis, and D. Varoutas, On the combination of exponential smoothing and diffusion forecasts: an application to broadband diffusion in the OECD area, Technological Forecasting and Social Change, vol. 78, no. 1, pp. 163 17, 211. 11 N. Meade and T. Islam, Forecasting with growth curves: an empirical comparison, International Journal of Forecasting, vol. 11, no. 2, pp. 199 21, 199. 12 K. Li, Z. Chen, Y. Li, and A. Zhou, An application of genetic programming to economic forecasting, in Current Trends in High Performance Computing and Its Applications, Proceedings of the International Conference on High Performance Computing and Applications, Part I, pp. 71 8, Springer, 2. 13 OECD Broadband portal, 211, http://www.oecd.org/dataoecd/22/12/3974779.xls. 14 OECD Broadband portal, 211, http://www.oecd.org/dataoecd/22/13/3974788.xls. 1 M. A. Kaboudan, Forecasting with computer-evolved model specifications: a genetic programming application, Computers and Operations Research, vol. 3, no. 11, pp. 1661 1681, 23.
Advances in Operations Research Volume 214 Advances in Decision Sciences Volume 214 Journal of Applied Mathematics Algebra Volume 214 Journal of Probability and Statistics Volume 214 The Scientific World Journal Volume 214 International Journal of Differential Equations Volume 214 Volume 214 Submit your manuscripts at International Journal of Advances in Combinatorics Mathematical Physics Volume 214 Journal of Complex Analysis Volume 214 International Journal of Mathematics and Mathematical Sciences Mathematical Problems in Engineering Journal of Mathematics Volume 214 Volume 214 Volume 214 Volume 214 Discrete Mathematics Journal of Volume 214 Discrete Dynamics in Nature and Society Journal of Function Spaces Abstract and Applied Analysis Volume 214 Volume 214 Volume 214 International Journal of Journal of Stochastic Analysis Optimization Volume 214 Volume 214