Combined artificial bee colony algorithm and machine learning techniques for prediction of online consumer repurchase intention

Abstract

A novel paradigm in the service sector i.e. services through the web is a progressive mechanism for rendering offerings over diverse environments. Internet provides huge opportunities for companies to provide personalized online services to their customers. But prompt novel web services introduction may unfavorably affect the quality and user gratification. Subsequently, prediction of the consumer intention is of supreme importance in selecting the web services for an application. The aim of study is to predict online consumer repurchase intention and to achieve this objective a hybrid approach which a combination of machine learning techniques and Artificial Bee Colony (ABC) algorithm has been used. The study is divided into three phases. Initially, shopping mall and consumer characteristic's for repurchase intention has been identified through extensive literature review. Secondly, ABC has been used to determine the feature selection of consumers' characteristics and shopping malls' attributes (with > 0.1 threshold value) for the prediction model. Finally, validation using K-fold cross has been employed to measure the best classification model robustness. The classification models viz., Decision Trees (C5.0), AdaBoost, Random Forest (RF), Support Vector Machine (SVM) and Neural Network (NN), are utilized for prediction of consumer purchase intention. Performance evaluation of identified models on training-testing partitions (70-30%) of the data set, shows that AdaBoost method outperforms other classification models with sensitivity and accuracy of 0.95 and 97.58% respectively, on testing data set. This study is a revolutionary attempt that considers both, shopping mall and consumer characteristics in examine the consumer purchase intention.

Log-Based Session Profiling and Online Behavioral Prediction in E–Commerce Websites

“…Some models address the challenge of distinguishing between buying and non-buying sessions (B/NB, two possible prediction outcomes) [4], [7], [8], [29]- [35]. Alternatively, other works concentrate on calculating the probability that a customer buys either a specific product (B-Prod) [20], [36]- [38] or a class of products (B-CProd) [39], [40], makes a purchase in the next visit to the online store (Next) [41]- [43], or repurchases in a future session (ReP) [44], [45]. Time constraints have also been considered as a part of some prediction models to estimate the purchasing probability of a user for the next day (Next-D) [46], for the next year (Next-Y) [47], or over time (NoT) [40].…”

“…These data sources are processed to discover and select the features/attributes that will be used to create prediction models. As an exception, [34], [45], [47] propose the use of questionnaires for gathering information regarding customers' preferences and behaviors in the hiring of (banking) services.…”

“…As shown in Table 1, the most common attributes are customers' personal (P) or demographic (D) data, product interest scores (PI), customers' navigation (NB) or purchasing behaviors (PB), or historical purchasing data (HP) (the RFM value or payments, for instance). Nevertheless, some proposals select alternative interesting attributes, such as the use of shopping carts (SC) [32], [43], seller's reputations and facilities (SRF) [45], customers' opinions (CO) [47], changes in user behavior (ChB) [46] or interactions of users with Web pages and their elements (Int) [7]. From a methodological point of view, two approaches should be emphasized.…”

“…Among the classification methods, the most common are Neuronal Networks (NN), Decision Trees (DT), Support Vector Machines (SVM), Random Forest (RF) and Naives-Bayes models (NBM). The approaches based on these methods make their predictions using a single classifier (S-Class, [33], [34], [43], [46]) or by combining multiple classifiers to improve the accuracy of the results (M-Class, [7], [8], [29]- [31], [40], [42], [44], [45]). In the last case, a combination algorithm integrating the predictions of the different classifiers is needed.…”

Churn prediction in Turkey's telecommunications sector: A proposed multiobjective–cost‐sensitive ant colony optimization

“…Elapsed time (Keramati et al, 2014) Elapsed time since last renewal 10. International plan yes/no (Burez & Van den Poel, 2009) Whether to use the international plan (D) Customer demographic characteristics (B. Huang et al, 2010;Hung et al, 2006;Keramati & Ardabili, 2011;N. Kim et al, 2012;Kumar, Kabra, Mussada, Dash, & Rana, 2019) 1 Customer's city (Karahoca & Karahoca, 2011) Where the customer lives. It is essential for churn analysis to distinguish whether the customer is living in rural or urban areas.…”