ANALYSING PERFORMANCE: A COMPARITIVE STUDY OF MACHINE LEARNING MODELS

Abstract

Sales forecasting is a fundamental company technique that influences inventory management, revenue estimation, and investment decisions. This study compares machine learning algorithms for improving sales projections, taking into account the financial implications of fluctuating sales seasons. Failure to foresee sales patterns can lead to significant losses, particularly for hug firms such as Walmart. The study makes use of a thoroughly pre-processed and merged dataset that includes features such as store details, department temperature, fuel price, markdowns, and the consumer price index (CPI), as well as five markdown features. Data uniformity is emphasized to assure model effectiveness by assigning equal weight to each feature and supporting a variety of techniques. Five machine learning models are used for comparative study. Random forest outperforms the others, obtaining a sales forecasting. The study underlines the importance of data standardization and preprocessing in model performance and offers guidance on model selection for sales forecasting. To summarize, this study contributes to the field of sales forecasting and offers practical advice for firms looking for effective machine learning models for strategic decision-making. The findings provide useful guidelines for using predictive analytics in the retail industry, emphasizing the necessity of recognizing and adjusting to sales changes for long-term economic success.