Abstract

This project aims to forecast gold prices using various time series analysis techniques. The dataset consists of monthly gold prices, which are analyzed using statistical and machine learning models to generate future predictions. The methodology includes data preprocessing, visualization, model training, and evaluation using key performance metrics. The models implemented include Linear Regression, Naive Forecasting, and Exponential Smoothing. The project offers insights into the effectiveness of various forecasting techniques and aids in identifying trends and seasonal patterns in gold prices.

Methodology

Data Preprocessing

The dataset containing monthly gold prices is loaded from a CSV file.

Missing values are handled, and the date column is converted into a proper datetime format.

The dataset is structured as a time series with appropriate indexing.

Exploratory Data Analysis (EDA)

A Time Series Plot is created to observe overall trends.

Box Plots are used to analyze gold price distributions across different years and months.

A Seasonal Plot highlights recurring patterns in gold prices.

Model Implementation

Naive Forecasting: Uses the most recent observed value as the forecast for all future points.

Linear Regression: Fits a trend line to the training data to capture the overall direction of price movements.

Exponential Smoothing: Applies a smoothing technique with an additive trend and seasonality to generate accurate forecasts.

Model Evaluation

The Mean Absolute Percentage Error (MAPE) is used to measure the accuracy of each model.

Forecast plots compare actual vs. predicted values, including confidence intervals.

Results

Linear Regression provided a general trend but struggled to capture seasonality.

Naive Forecasting was simple but performed poorly for long-term predictions.

Exponential Smoothing captured both trend and seasonal components effectively, resulting in more accurate forecasts.

The model performances were assessed using MAPE, with Exponential Smoothing achieving the best results.

The findings suggest that Exponential Smoothing is the most effective technique for forecasting gold prices due to its ability to capture both trend and seasonal variations.