GPU Implementation of Sales Forecasting with Linear Regression

Abstract

Forecasting of sales is very important in any business as it helps managers to learn from historical data and make informed decisions. This generally involves intensive processes using spreadsheets which require inputs from all levels in an organization. This approach introduces bias and is generally not accurate. There are several methods that have been used in the past to forecast sales, such as Exponential Smoothing, Moving Average, and Autoregressive Moving Average (ARMA). Due to the nature of the data, it usually takes more time for these methods to analyze the sales data and make predictions. In this paper, the sales data is analyzed and predictions are made by using linear regression as implemented on the GPU to make the process faster. Sales forecasting is made possible by finding best fit line by linear regression techniques (i.e. linear convolution). To illustrate this process, simulated sales data was used. The sales forecasting with linear regression implementation using GPU was compared to the CPU implementation and a speedup of up to 7.557x was achieved.

Design and Implementation of Sales Prediction Model Using Decision Tree Regressor over Linear Regression Towards Increase in Accuracy of Prediction

“…Researchers from a spread of fields such as mathematics, system learning, sample recognition, and statistics mining have looked at the hassle of building a decision-making answer from to be had records. A paper by [8]Ayomide Yusuf (Yusuf and Alawneh 2018) shows a new way of forecasting sales with integration of GPU with linear regression. (Grąbczewski 2014) found that tests performed to study management features in Decision Tree and import controls are inconsistent [9].…”