In the article, a modification of the new criterion of the adequacy of regression equations introduced by the author-the "consistency of behavior" criterion or the so-called CB-criterion is introduced. It is based, unlike the traditional adequacy criteria in the regression analysis, not on the analysis of the approximation errors of the equation, but on the correlation of the signs of the increments of the actual and calculated values of the dependent variable of the equation by the observation numbers. Therefore, even for equations with high values of classical verification criteria, the CB criterion may have low significance. In contrast to the SP criterion, the generalized criterion for consistency of behavior (GCB criterion) proposed in this article assumes the correlation of the indicated increments for pairs of observations with arbitrary numbers, which makes it possible to reveal the complete picture in accordance with the behavior of the actual and calculated values of the dependent variable of the equation throughout the sample, all possible cross-links. In addition, the paper proposes an algorithm for maximizing the value of an GCB test with a fixed or slightly degraded value chosen by the researcher for the loss function as a sum of the absolute values of the approximation errors corresponding to the Manhattan distance or the method of the smallest modules. This algorithm allows us to reduce this problem to the problem of partially-boolean linear programming of low dimensionality. It also provides for the possibility of combining the GCB criterion with the loss function by forming their linear convolution. In this case, it is possible to give each of its components a different weight, depending on which criterion the decision-maker considers more or less important. With the software implementation of this algorithm, the LPsolve program can be effectively used on the Internet.
1. Mudrov VI, Kushko V.A. Methods for processing measurements. Quasi-like estimates .- M .: Radio and Communication, 1983.-304p.
2. Noskov S.I. The technology of modeling objects with unstable functioning and uncertainty in the data. Irkutsk: Oblinformpechat, 1996.-320 p.
3. Noskov S.I. Construction of econometric dependencies taking into account the criterion of "consistency of behavior" / / Cybernetics and system analysis.-1994.-No.1-P.177-180.
4. Noskov S.I. The criterion of "consistency of behavior" in regression analysis // Modern technologies. System analysis. Modeling.-2013.-No.1-P.107-111.
5. Jingfei Yang M. Sc. Power System Short-term Load Forecasting: Thesis for Ph.d degree. Germany, Darmstadt, Elektrotechnik und Informationstechnik der Technischen Universitat, 2006. 139 p.
6. Day-Ahead Electricity Price Forecasting Using the Wavelet Transform and ARIMA Models / A.J. Conejo [at al.] // IEEE transaction on power systems. 2005, Vol. 20, No. 2. P.P. 1035 – 1042.
7. Armstrong J.S. Forecasting for Marketing // Quantitative Methods in Marketing. London: International Thompson Business Press, 1999. P.P. 92 – 119.
8. Draper N., Smith H. Applied regression analysis. New York: Wiley, In press, 1981. 693 p.
9. Pradhan R.P., Kumar R. Forecasting Exchange Rate in India: An Application of Artificial Neural Network Model // Journal of Mathematics Research. 2010, Vol. 2, No. 4. P.P. 111 – 117.
10. Yildiz B., Yalama A.,Coskun M. Forecasting the Istanbul Stock Exchange National 100 Index Using an Artificial Neural Network // An International Journal of Science, Engineering and Technology. 2008, Vol. 46. P.P.36 – 39.
11. Zhu J., Hong J., Hughes J.G. Using Markov Chains for Link Prediction in Adaptive Web Sites // 1st International Conference on Computing in an Imperfect World, UK, London, 2002. P.P. 60 – 73.
12. Singh S. Pattern Modelling in Time-Series Forecasting // Cybernetics and Systems-AnInternational Journal. 2000, Vol. 31, No. 1. P.P. 49 – 65.