Implementing And Comparing Forecasting Algorithms For Sales Data Using Python: Facebook Prophet, Sarima And Holt-winters

The main aim was to illustrate the difference between the new age machine learning algorithm known as Facebook Prophet to conventional statistical models such as SARIMA and Holt Winters. In this research, time-series sales data were used, although the research recognized seasonality as a major factor that affected the sales. A Python script has been utilised as the main application and allowed for the assessment of multiple algorithms of forecasting. This compared Facebook Prophet, SARIMA and Holt Winters on the basis of performance, stability, and suitability on the data. This involves coding every algorithm in the script and evaluating how effective they are each could be as measured by factors like the R-squared coefficient. The evaluation pointed out the fact that Facebook Prophet was identified as the most suitable model in respect to the selected dataset and error measures. It has achieved an R-squared of 0. 94 for fitted data which revealed it as the best result brought by using simulated annealing when fine tuning the sales data, which underlined the programmes capacity to portray the sales data patterns accurately. In using Holt-Winters much of was learnt although it had an R-value squared of 0. Elaborated 87 (used simulated Annealing in fine tuning) and 0. 87 (used simulated annealing in fine tuning) on the training set and one would like to point out that such performance does not remain constant and depends upon specific data and specified measurement of performance. However, for the rest of the algorithms, the performance was more or less almost equivalent with only a slight leave behind by SARIMA scoring a 0. 61 (the best result by such fine tuning obtained by simulated annealing). This stresses the significance of the decision of which algorithm to use in specific conditions and with specific data. The second part of the thesis focuses on the functionality of the Python scripts, that was implemented during the work on the project. Each of the scripts is profusely commented and explained in detail with pictures and step-by-step instructions of the process of data analysis and data preprocessing. Seeing the outcomes at work, you will be able to comprehend how these scripts juxtaposes and rank these algorithms in accordance to specific measures such as, R-squared.The research underscores the potential of ML in sales forecasting, particularly Facebook Prophet’s ability to deliver accurate predictions. Furthermore, the developed Python script offers a versatile tool for running and comparing multiple forecasting models simultaneously. This readily implementable solution worked with businesses across various sectors, from corporations managing inventory to logistics providers anticipating client needs, to leverage historical data and optimize their sales forecasting processes, ultimately contributing to improved efficiency and success.

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Read More