Application of Machine Learning Technique for Demand Forecasting: A Case Study of the Manufacturing Industry

The objective of this work is to develop a machine learning-based Support Vector Machine (SVM) demand forecasting model and its application in supply chain management. The proposed SVM model will predict future demand with high accuracy as compared to the conventional forecasting methods. To demonstrate the effectiveness of the present model, demand forecasting issue was investigated in a piston-manufacturing industry as a real-life case study. In this proposed research, an SVM model is developed using radial basis kernel function and sigmoid function to forecast monthly piston demand for Bajaj Discover motorbikes. Various factors that affect the product demand such as produced units, inventory, sales cost, and the number of competitors have been taken into consideration in the development of the model. A comparative analysis of the SVM model and various traditional forecasting methods used in the company like exponential smoothing, moving average, and autoregressive model has been done and the best demand forecasting model has been recommended to the case company.