学术文献库

Predicting future consumer behaviour is one of the most challenging problems for large scale retail firms. Accurate prediction of consumer purchase pattern enables better inventory planning and efficient personalized marketing strategies. Optimal inventory planning helps minimise instances of Out-of-stock/ Excess Inventory and, smart Personalized marketing strategy ensures smooth and delightful shopping experience. Consumer purchase prediction problem has generally been addressed by ML researchers in conventional manners, either through recommender systems or traditional ML approaches. Such modelling approaches do not generalise well in predicting consumer purchase pattern. In this paper, we present our study of consumer purchase behaviour, wherein, we establish a data-driven framework to predict whether a consumer is going to purchase an item within a certain time frame using e-commerce retail data. To model this relationship, we create a sequential time-series data for all relevant consumer-item combinations. We then build generalized non-linear models by generating features at the intersection of consumer, item, and time. We demonstrate robust performance by experimenting with different neural network architectures, ML models, and their combinations. We present the results of 60 modelling experiments with varying Hyperparameters along with Stacked Generalization ensemble and F1-Maximization framework. We then present the benefits that neural network architectures like Multi Layer Perceptron, Long Short Term Memory (LSTM), Temporal Convolutional Networks (TCN) and TCN-LSTM bring over ML models like Xgboost and RandomForest.