📈 FinanceTimeSeriesForecast Dashboard

Harikrishnan M | GEC Thrissur | B.Tech CSE Sem 4

🚀 Market Vibrations

This research project treats financial time series data (TCS Stock) as non-stationary signals. By applying Short-Time Fourier Transforms (STFT), we extract time-frequency features (spectrograms) which are then processed by a 2D Convolutional Neural Network (CNN).

This captures cyclic market behaviors and localized volatility spikes that traditional 1D LSTM models often overlook.

📊 Comparative Signal Analysis

Domain	Transformation Type	Resulting Visualization
Time	x[n] Normalized Price
Frequency	Global Fourier Transform (FFT)
Time-Frequency	STFT Spectrogram S(t, f)

🛠️ Terminal Setup & Execution

# 1. Environment Setup
git clone https://github.com/ha7-piixel/FinanceTimeSeriesForecast.git
pip install -r Requirements.txt

# 2. Execute Data & Training Pipeline
python3 src/generate_plots.py
python3 src/train.py

# 3. Final Report Compilation
cd results && pdflatex report.tex

🧠 CNN Layer Definition

🔹 Input: 2D Spectrogram Map (1, 33, 94)
🔹 Conv2D: 16 Kernels (Local Spatial Extraction)
🔹 MaxPool: Dimensionality Reduction
🔹 Dropout: Overfitting Prevention (0.25)
🔹 Dense: 64-unit Linear Fusion Layer