Initial Population Based Crayfish Optimization Algorithm with Bi-directional Long Short-Term Memory of Customer Churn Prediction

Churn prediction in the retail industry refers to the process of identifying customers who have stopped buying products from retailers by analysing patterns in customer behaviour and transaction data. The retail industry entails businesses where retailers sell their products directly to the customers. However, the challenge lies in the insufficient data available to predict churn, which makes it difficult to identify customers who have stopped buying from retailers. To overcome this problem, an Initial Population based Crayfish Optimization Algorithm (IP-COA) is proposed for feature selection and churner predict in retail industry, using a Bi-directional Long Short-Term Memory (LSTM). The data from Kaggle repository is used as input in this research. In the pre-processing stage, normalization techniques are employed to refine and resize the input data before feeding it into the feature selection process. The IP-COA helps identify the most significant features, while eliminating the unnecessary features. At last, the selected features are classified as churners and non-churners, helping retainers enhance their business strategies. The evaluation metrics used for the analysis of the proposed method are, accuracy, precision, recall and F1-score with the secured values being 98.9%, 98.0%, 98.3%, and 98.1%. This hence proves the IP-COA's supremacy over the existing GAT-CNN, RNN and DNN-BiLSTM.