KGHM Downside Risk Prediction

This project is a C# academic implementation of a decision support system for financial risk management in the mining sector. It aims to predict the probability of a bearish (downside) movement for the KGHM stock one day in advance, using market data and commodity prices.

Project Overview

The system processes raw historical market data, extracts relevant financial features, and performs a binary classification using a Logistic Regression model powered by the Accord.NET framework.

Key Features

• OOP Architecture: Modular design following strict SOLID principles, featuring interfaces, inheritance, and encapsulation.
• Data Preprocessing Pipeline:
  • Raw CSV data ingestion using custom IO components.
  • Dynamic feature engineering (Log-returns, Moving Averages).
  • Feature standardization (Z-Score) to ensure model stability.
• Machine Learning: Binary classification using Accord.Statistics.Models.Regression.LogisticRegression.
• Professional Standards: No hardcoded values, XML-documented code, and fully configurable parameters.

Architecture

The project is organized into logical layers:

• Models: Data representations (KghmRecord).
• IO: Data ingestion strategies (CsvReader, CsvColumnMapping).
• Processing: Mathematical feature extraction (FeatureExtractorBase, LogReturnExtractor, MovingAverageExtractor).
• ML: Training and evaluation logic using Accord.NET.

Requirements

• .NET 8.0 SDK
• Accord.NET Framework

Usage

1. Place your kghm.csv file in the appropriate directory.
2. Configure column mappings in Program.cs (or via CsvColumnMapping).
3. Run the project using:

   dotnet run