Abstract: Background: Accurate sales prediction in the retail sector, especially in the fast-paced fashion market, is crucial for optimizing inventory, reducing costs, and avoiding out-of-stock situations. Retailers and wholesalers face significant challenges in forecasting future sales and understanding market trends, both of which are essential for effective pricing strategies. Methods: This study compares several machine learning and deep learning techniques to forecast sales in the e-commerce fashion retail industry. The models evaluated include Linear Regression, Polynomial Regression, Decision Tree (DT), Support Vector Machine (SVM), XGBoost, Long Short-Term Memory (LSTM), and a hybrid CNN-LSTM framework. The hybrid CNN-LSTM approach leverages convolutional networks for identifying features and recurrent layers to model sequential patterns over time. The models' performances are assessed using metrics like R2 score, RMSE, MAE, and MAPE. Findings: The research reveals that the CNN-LSTM hybrid model significantly outperforms the others in terms of accuracy and robustness, making it the most effective for predicting sales in the fashion retail sector. Novelty and Applications: This study introduces a novel application of the CNN-LSTM hybrid model for sales prediction in the e-commerce fashion retail industry. The integration of convolutional and recurrent neural networks enables the model to effectively handle the intricacies of sales data, combining short-term feature extraction with long-term trend analysis. The superior performance of this model provides a valuable tool for retailers, helping them to predict sales more accurately and optimize product pricing based on anticipated sales. This approach offers a significant advancement over traditional sales prediction methods, contributing to more informed and strategic decision-making in the retail industry.

Keywords: Time Series forecasting, Sales forecasting, LSTM (Long Short-Term Memory), CNN – LSTM (Convolutional Neural Network- Long Short-Term Memory Network), Hybrid Machine Learning, DT (Decision Tree), XGBoost, SVM (Support Vector Machine) algorithm; supervised machine learning techniques.

| DOI: 10.17148/IARJSET.2024.111113