Heuristic approach to online purchase prediction based on internet store visitors classification using data mining methods

Abstract

Last years research gave some preliminary results in approaches to customer online purchase prediction. However, it still remains unclear what exact set of features of data instances should be incorporated in a model and is enough for prediction, what is the best data mining method (algorithm) to use, how stable over time could be such a model, whether a model is transferable from one online store to another. This study is focused on a heuristic approach to dealing with the problem under conditions of such theoretical and methodological diversity in order to find a quick and inexpensive first approximation to the solution or at least to find useful patterns and facts in the data.

A new fuzzy rule-based optimization approach for predicting the user behaviour classification in M-commerce

“…It begins along with a first populace of arbitrarily or even heuristically produced people, and also breakthroughs towards far better people through administering hereditary drivers created on the hereditary procedures developing in mother of nature. Parkhimenka et al [29] to develop a brand-new populace, people are actually picked depending on to their exercise. An achievable service in the health and fitness feature takes on a decoded chromosome as input, and as well as creates an unbiased worth of functionality to the input chromosome.…”

Using Word2Vec Recommendation for Improved Purchase Prediction

“…On the other hand, in some works, features are learned as a consequence of the click actions using Recurrent Neural Network (RNN) methods [37], [38]. Moreover, in [32], heuristic methods were utilised to have quick coverage by estimating the initial seeds for prediction models. Experiment results show that they were able to get 99% accuracy on purchase intention detection in sessions.…”