Sales Analysis and Forecasting: Time Series Modeling for Business Decision Making

This thesis investigates the analysis and modeling of a monthly revenue time series with the objective of understanding its underlying dynamics and improving forecasting accuracy. Revenue time series are typically characterized by trends, seasonal patterns, and irregular fluctuations, which make their modeling both challenging and essential for effective business decision-making. The analysis is based on ARIMA-type models, whose main properties are introduced and discussed. Model performance is assessed through appropriate statistical metrics and diagnostic tests. The study also provides a detailed description of the dataset, including data extraction, preprocessing, and transformation procedures required to obtain a reliable series for analysis. An exploratory data analysis is first conducted, involving visual inspection, descriptive statistics, and the identification and treatment of outliers and seasonal components. Building on these findings, both non-seasonal and seasonal time series models are estimated and compared in terms of their ability to capture the observed dynamics and produce accurate forecasts. Overall, the proposed methodology offers a structured and replicable framework for the analysis and forecasting of monthly revenue time series, supporting data-driven decision-making in economic and business applications.