🧠 ML Final Project: Predicting Small-Cap Stock Movements

📈 Goal

Use machine learning models to predict whether a small-cap stock (from the Russell 2000 Index) will go up or down tomorrow based on historical technical indicators.

🗂️ Dataset

• Source: Yahoo Finance via yfinance
• Stocks: 200 randomly selected small-cap stocks from the Russell 2000
• Period: 2015–2025 (daily prices)
• Target: Binary — 1 if tomorrow's closing price > today's, else 0

🛠️ Features Used

• SMA_7, SMA_14 — short-term moving averages
• SMA_50, SMA_200 — medium/long-term moving averages
• RSI_14 — Relative Strength Index
• MACD, MACD_Signal, MACD_Hist — momentum indicators
• Volatility — price variation metric

🤖 Models Implemented

• Logistic Regression – Simple linear baseline
• AdaBoost – Boosted weak learners
• K-Means Clustering – Used for behavioral grouping (unsupervised)
• Random Forest – Ensemble of decision trees
• Baseline Model – "Predict same as yesterday" rule

⚙️ Accuracy Results

Model	Accuracy
Logistic	~51.5%
AdaBoost	~51.7%
Random Forest	~51.8%
Baseline (naive)	~60.0%

⚠️ The rule-based baseline surprisingly performed better than ML models — highlighting the noisy, random nature of short-term market predictions.

📊 Visualizations

• Accuracy comparison bar chart
• Confusion matrices
• Random Forest feature importance chart
• Stock price + SMA trendline example

All visuals saved under /visuals/.