Augmented Gold Price Forecasting via Image Processing and Machine Learning Fusion

This project presents a comprehensive approach to analyze and estimate gold prices by integrating image processing, time series forecasting, and machine learning techniques. The workflow encompasses multiple stages, starting with data preparation and image preprocessing using Tensor Flow, Keras, and OpenCV. Image data, including width, height, color distribution and brightness/contrast, is explored and visualized, while financial data is loaded and analyzed for descriptive statistics and time series patterns. The core of the project lies in the development of a Convolutional Neural Network (CNN) model for binary classification of gold and non-gold images. Various optimizers are experimented with to enhance model performance. The trained models are saved and evaluated on new images, incorporating a unique analysis of gold quality based on color, edge density, and texture. For time series forecasting, Autoregressive Integrated Moving Average (ARIMA) models are employed to predict gold prices in different countries. The project also explores the use of Tensor Flow Transform (TFT) for data preprocessing, enhancing the robustness of the analysis. In addition to gold price prediction, the project introduces a Stream lit web application for interactive image quality analysis. Users can upload gold images, assess quality attributes, adjust predicted gold prices based on quality, and forecast future prices for selected countries within a specified time range. The interdisciplinary nature of this project, combining computer vision, time series analysis, and machine learning, offers a holistic approach to understanding and estimating gold prices. The methodology and tools employed make this project valuable for researchers, analysts, and practitioners seeking a multifaceted perspective on gold market dynamics.