TrendCast

Abstract

Demand forecasting plays a crucial role in strategic planning and decision-making across multiple industries, 
including retail, food services, dairy, supermarkets, and electronics. This project presents a multi-domain demand 
forecasting system designed to predict future product demand using historical time-series data. 
The system processes structured datasets containing Date, Product, and Demand attributes. Data preprocessing 
is performed to clean and aggregate the data on a monthly basis, enabling better identification of long-term 
trends, seasonality, and demand patterns. Forecasting is carried out using advanced statistical and machine 
learning models such as ARIMA (AutoRegressive Integrated Moving Average) and Facebook Prophet, both of 
which are effective for time-series analysis. 
The dataset is split into training and testing sets (80:20 ratio) to ensure model validation and reliability. Model 
performance is evaluated using standard error metrics, including Mean Absolute Error (MAE), Root Mean 
Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). The model with the best performance is 
selected for generating future demand predictions. 
The system also provides a user-friendly interface that allows users to upload datasets, select specific products, 
and define the forecast horizon. The output includes forecasted demand values, visual graphs, and downloadable 
CSV reports. 
This project delivers a scalable and efficient solution for inventory management, demand planning, and business 
decision-making across diverse domains.