Leveraging CNN-LSTM Model for Adaptive Pricing in Fashion Retail Segment

Abstract

ntroduction: Pricing optimization in the fashion retail sector can be very complex, thus necessitating correct forecasting by sales and adjustable pricing systems. Traditional pricing models usually do not consider the dynamic market conditions. Hence, advanced machine learning techniques must be employed to support effective decision-making. Herein, we directly propose a learning machine-based framework for sales prediction and price optimization to boost market competitiveness with higher profitability.

Objectives: The primary objectives of this study are:

1. To develop a robust machine learning-based sales prediction model for fashion retail.


2. To implement an optimized pricing strategy using deep learning techniques.


3. To maximize profit through a data-driven pricing approach.n.

Methods: A two-phase approach was employed in this research. The first phase involved sales prediction using various machine learning models, including Long Short-Term Memory (LSTM), Decision Trees, XGBoost, Support Vector Regression (SVR), Polynomial Regression, and Linear Regression. The most effective model, LSTM, was selected for further integration. In the second phase, a CNN+LSTM hybrid model was used for price optimization. CNN was utilized to extract features from historical sales data over a 28-day context window, and the LSTM model predicted demand for the next 14 days. A profit matrix was generated across 15 price levels per SKU, enabling optimal pricing selection.

Results: LSTM was much more accurate than other machine-learning methods in demand forecasting. The proposed hybrid CNN+LSTM model has been effectively used in detecting optimal price points that analyze historical sales patterns and maximize profitability. The model found price adjustments leading to better daily and total 15-day profits and thus enabling the retailers to make real-time data-based pricing decisions.

Conclusions: This study demonstrates the effectiveness of integrating deep learning and machine learning models for pricing optimization in fashion retail. The proposed framework provides a systematic approach to predicting sales and setting optimal price levels, ultimately enhancing profitability and competitive positioning in the market. The research highlights the potential of AI-driven pricing strategies in retail operations, paving the way for future advancements in automated pricing mechanisms