Restaurant Revenue Prediction Applying Supervised Learning Methods

Authors

Md Yasin Ali Parh, Department of Bioinformatics and Biostatistics, University of Louisville, US\\ Department of Statistics, Islamic University, Kushtia, BangladeshFollow
Mst Sharmin Akter Sumy, Department of Bioinformatics and Biostatistics, University of Louisville, US\\ Department of Statistics, Islamic University, Kushtia, BangladeshFollow
Most Sifat Muntaha Soni, Department of Statistics, Islamic University, Kushtia, BangladeshFollow

Abstract

In the competitive world, it is difficult to make a decision where to open a restaurant outlet that produces maximum revenue. Especially, it is difficult to accurately extrapolate across geographies and culture based on the personal judgement and experiences. Supervised learning approach may play a vital role to determine the feasibility of a new outlet with the prediction of revenue. The goal of this study was to predict restaurant revenue of 100,000 regional tab food investment (TFI) restaurant locations across Turkey. Several supervised learning techniques were used to select the optimal model for prediction. The LASSO method was selected as the best supervised method for the prediction of revenue as determined by lowest test error. Other models were employed, but LASSO outperformed all other models and had the added benefit of simplicity and interpretability. The LASSO model was used to predict the revenue of 100,000 new restaurant site locations based on the coefficients termed using the training data.

Digital Object Identifier (DOI)

http://dx.doi.org/10.18576/jsap/120101

Recommended Citation

Yasin Ali Parh, Md; Sharmin Akter Sumy, Mst; and Sifat Muntaha Soni, Most (2023) "Restaurant Revenue Prediction Applying Supervised Learning Methods," Journal of Statistics Applications & Probability: Vol. 12: Iss. 1, Article 1.
DOI: http://dx.doi.org/10.18576/jsap/120101
Available at: https://digitalcommons.aaru.edu.jo/jsap/vol12/iss1/1