Shock absorber sales forecast with artifical neural networks for the Covid 19 process

Keywords: Shock Absorber Sales, WEKA, MatLab, ANN and Forecast Parameters, Covid 19 Today, many companies managed with an agile approach have to evaluate their data correctly and use it as an additional resource in the decision-making process. With the Covid-19 pandemic, the automotive industry has been adversely affected, as in many other sectors. In this study, an estimate of the sales of a shock absorber company for March, April and May 2020, which is the early pandemic period for Turkey, has been made and the negative impact of the pandemic has been tried to be reflected by comparing the difference between the estimated values obtained and the actual sales. ANN structures, estimation parameters and model trainings of WEKA and MatLab programs were used by obtaining monthly sales data from January 2016 to February 2020 from a shock absorber company. On the other hand, the actual 3 months sales figures covering the pandemic period, were taken from the company and named as actual sales and compared with the sales forecast results of WEKA and MatLab. As a result, it was observed that the sales of shock absorbers had an increasing trend before pandemic period but when the forecasts for the pandemic period and actual sales were compared, the sales of shock absorbers were adversely affected by the pandemic. According to the findings obtained in the research, it was concluded that the WEKA structure has larger margin of error in the estimation results and MatLab, which includes more ANNs in the established model, produces more successful predictions. However, it was predicted that the main reason for the high error rates of both programs for the predicted pandemic period within the scope of an estimation study was that there was no definition of crisis periods in the programs, and therefore the effect could not be reflected on the established ANN models.