Publication Date

Spring 2024

Degree Type

Master's Project

Degree Name

Master of Science in Computer Science (MSCS)

Department

Computer Science

First Advisor

Robert Chun

Second Advisor

William Andreopoulos

Third Advisor

Chandra Pavan Reddy Chada

Keywords

Dynamic forecasting engine, Seasonal ARIMA, XGBoost, LSTM, real time adjustments, personalization, tailored recommendations, data-driven dining.

Abstract

This project introduces a "dynamic forecasting engine," designed to transform the way restaurants predict sales. The engine dynamically handles seasonal ARIMA_HoltWinter hybrid model, XGBoost, and LSTM algorithms to dynamically select the best forecasting method based on data volume, variety, and customer taste preferences delving upon the spice level categorical sales. This guide differs from traditional crystal ball approaches because it has the ability to improve over time as new data comes in terms of spice levels. It emphasizes the importance of dataset size in the selection of machine learning algorithms through complexity for large datasets and simplicity for smaller ones along with keeping track of spice level sales modulation. Customer orders, tastes, and periodical variations are some of the key aspects that would help create an accurate forecast of this engine. Also integrated into this engine is a visual application that provides a complete dashboard for inventory management efficiency during operations planning, as well as personalization features that use trends related to customer preferences and past experiences to generate specific recommendations (Tavares et al., 2017). What if there was a virtual assistant that predicted your next orders based on your own preferences, making every dining experience unique? This paper looks at the future of data driven dining, where the "dynamic forecasting engine" becomes more than just a tool, but a companion to restaurants.

Download

Available for download on Thursday, May 22, 2025

Share

COinS