Forecasting Product Demand from Historical Data Using Machine Learning Methods

---

Project Team

---

Project

---

Project Video

---

Project Abstract

Recently, time series forecasting has become a subject of interest for many data scientists and
researchers alike. Both statistical and machine learning methods have been applied to time series
data in order to predict future phenomena like the weather and stock prices. In our global
climate, one avenue of time series forecasting that continues to become more imperative is the
prediction of the demand for perishable goods. Each year, vast quantities of food are wasted
globally with the number continuing to grow. To abate this crisis, an accurate time series model
could be used to predict the number of resources needed at the store level. However, current
forecasting models are plagued with inaccuracies and are only utilized as decision-making tools
for upper-level management who ultimately determine the final decision. We propose an
exploratory approach by evaluating different models on their ability to predict market demand.
As a baseline, a statistical model is included to measure if the machine learning models show
superior predictive capability. We observed that the machine learning models performed slightly
better than the statistical baseline. This shows that machine learning has the capability of better
predicting market demand than statistical models.

---

Evaluate this Project

Use this form link to provide feedback to the presenters, and add your project evaluation for award(s) consideration.