Financial Markets Analysis

This repo contains:

• Financial_Markets_Analysis.ipynb: notebook-based analysis
• nse_sensex (1).csv: historical daily stock data (NSE-style)
• frontend/: interactive dashboard (upload + validation + charts)
• backend/: FastAPI service for validation, analytics, and ML inference
• ml/: time-series-safe training and inference CLI

Architecture (local-first)

• Frontend (frontend/app.py): Streamlit UI for upload, analytics, and prediction.
• Backend (backend/app/main.py): FastAPI endpoints for dataset validation/storage, analytics, and /ml/predict.
• Local storage:
  • datasets: data/datasets/<dataset_id>/data.parquet + meta.json
  • model runs: artifacts/<run_id>/model.joblib + run.json
• ML (ml/fma_ml): leakage-safe feature pipeline, walk-forward evaluation, saved artifacts.

Dataset format (supported by the dashboard)

Required columns:

• DATE (parseable date)
• SYMBOL (ticker)
• CLOSE (numeric)

Recommended columns:

• OPEN, HIGH, LOW, VOLUME

The dashboard normalizes these to: date, symbol, close, open, high, low, volume.

Run the dashboard (Windows PowerShell)

Install Python 3.10 (user scope, secure source):

winget install --id Python.Python.3.10 -e --scope user
py -0p

Start the app:

cd C:\Users\.\Financial-Markets-Analysis
.\scripts\run_frontend.ps1

What a user does (local-first)

• One command (recommended):

cd C:\Users\.\Financial-Markets-Analysis
.\scripts\run_all.ps1

• Run the backend in one terminal:

cd C:\Users\.\Financial-Markets-Analysis
py -3.10 -m uvicorn backend.app.main:app --host 127.0.0.1 --port 8000

• Run the frontend in another terminal:

cd C:\Users\.\Financial-Markets-Analysis
.\scripts\run_frontend.ps1

• Open http://localhost:8501 and:
  • explore analytics on the repo dataset or upload a CSV
  • (optional) train a model via CLI and use the Prediction tab

Deployment note

This project is intentionally local-first.

If your machine policy blocks pip --user, use a single repo venv instead:

cd C:\Users\.\Financial-Markets-Analysis
py -3.11 -m venv .venv
.\.venv\Scripts\python -m pip install -r .\frontend\requirements.txt
.\.venv\Scripts\python -m streamlit run .\frontend\app.py

ML (local, time-series-safe)

Install ML deps:

cd C:\Users\.\Financial-Markets-Analysis
py -3.10 -m pip install -r .\ml\requirements.txt

Train a baseline next-day direction model using a saved dataset:

$DATASET_ID="46648df56102493a9b366b91862fedae"
py -3.10 -m ml.fma_ml.cli train --dataset-id $DATASET_ID --symbol ASIANPAINT --model logreg --splits 5 --out artifacts

The command prints the run folder path (contains model.joblib + run.json).

Predict next-day direction probability using a saved run:

$RUN_ID="85181c56309a4e62a2eb96f60ad0371e"
$DATASET_ID="46648df56102493a9b366b91862fedae"
py -3.10 -m ml.fma_ml.cli predict --run-id $RUN_ID --dataset-id $DATASET_ID --symbol ASIANPAINT