Fog Computing Enabled Locality Based Product Demand Prediction and Decision Making Using Reinforcement Learning

Wastage of perishable and non-perishable products due to manual monitoring in shopping malls creates huge revenue loss supermarket industry. Besides, internal external factors such as calendar events weather condition contribute excess wastage different regions supermarket. It is a challenging job know about the manually supermarkets region-wise. Therefore, management needs take appropriate decision action prevent products. The fog computing data centers located each region can collect, process analyze for demand prediction making. In this paper, product-demand model designed using integrated Principal Component Analysis (PCA) K-means Unsupervised Learning (UL) algorithms making developed State-Action-Reward-State-Action (SARSA) Reinforcement (RL) algorithm. Our proposed method cluster into low, medium, high-demand product by learning from features. Taking derived model, distributing low-demand be made SARSA. Experimental results show that our datasets well with Silhouette score ≥60%. adopted SARSA-based outperforms over Q-Learning, Monte-Carlo, Deep Q-Network (DQN), Actor-Critic terms maximum cumulative reward, average reward execution time.