Safat Mahmood, Prairie View A&M University

Demand forecasting using convolutional neural networks (CNN) on big real world data

Abstract: Demand forecasting has become very significant in recent years because of its accurate and efficient resource allocation for companies. However, demand forecasting usually involves processing very large and sophisticated data sets and it is very challenging to achieve accurate prediction. Although there are many existing methods in the literature, they could not be directly applied to big data. In this presentation, a novel demand forecasting method using deep learning specifically, a Convolutional Neural Network (CNN) based model is proposed to process huge amounts of data. A unique feature of the proposed method is that it has adaptable computational intricacy even with the increasing dimensionality of the data. As a result, the proposed method has excellent scalability. Large data sets, namely the usage data of Bike Sharing Service in New York City are used to validate the proposed method. The goal is to predict hourly bike rental demand in every station of New York City to improve the service and environment, and eventually build a “Green” city. The results demonstrate the effectiveness and superior performance of the proposed scheme.

Presentation Author(s):
Safat Mahmood*, Dr. Lijun Qian, Dr. Xishuang Dong

Judging Forms Official judges only