Selection behavior preference subdivision algorithm

Abstract

The invention relates to a selection behavior preference subdivision algorithm. According to the algorithm, model parameters are estimated on the basis of selection behavior data and a discrete selection model so as to serve as representations of selection behavior preferences. The algorithm comprises the following steps of: decomposing the data into a plurality of sub-samples which are solvable to the lowest limit by using a Monte Carlo algorithm, wherein basis of the decomposition is a similarity between the model parameters, namely, comprehensively considering a relative difference betweentwo model parameters and credibility of the difference; classifying all the sub-samples step by step which are solvable to the lowest limit by taking the similarity as a distance index by using a hierarchical clustering algorithm; and finally determining a proper quantity of models according to a Bayesian information standard so as to obtain subdivided selection behavior preference types. According to the algorithm provided by the invention, heterogeneity in selection behavior data is mined; and compared with basic discrete selection model algorithm, the algorithm is capable of obtaining moreaccurate selection behavior preferences and models of different types of persons or groups, and laying a foundation for obtaining more accurate and pointed behavior prediction results for related applications.

Description

technical field [0001] The invention relates to an algorithm for subdividing selection behavior preferences, which is applied to many fields involving individual or collective selection and decision-making, such as retail business, transportation and travel, and enterprise management. Background technique [0002] In the context of increasingly abundant data resources, many industries increasingly need to collect personal behavior data to predict human behavior so as to make the industry develop better, such as shopping behavior and travel behavior. The choice behavior preference analysis method has been widely used. By collecting the choice behavior data of individuals or organizations, a behavior preference model is constructed, and the decision-making process is simulated to realize behavior prediction. [0003] Multinomial Logit Model is the most commonly used behavioral preference model, its mathematical form is concise, and it is easy to implement and apply. The model...

AI Technical Summary

Problems solved by technology

However, in reality, there are varying degrees of heterogeneity in individual behavior and preferences, and it is difficult to grasp a single model in a complete and precise manner, resulting in deviations in understanding preferences and behavior predictions, and bringing negative effects to industry decisions

Existing methods to solve this heterogeneity, either subjectively classify the population and establish a corresponding model, or mathematically improve the model structure, but the applicability is poor due to the increase of model complexity

Images