Time Series Analysis Web App

A web application for exploratory time series analysis, built for a KNTU course project. Users upload a CSV/XLSX file (or generate a synthetic series) and get autocorrelation (ACF), partial autocorrelation (PACF), trend, differencing, MA-model parameter estimation, and AI-generated (Persian) commentary on the results.

The UI is in Persian (RTL). The frontend is a React + TypeScript SPA; all numerical analysis runs in a Python/Flask backend.

Features

• CSV/XLSX upload & parsing into a labeled time series.
• Autocorrelation (ACF) for lags 1–10, plus trend coefficient estimation via linear regression.
• Partial autocorrelation (PACF) (via statsmodels) with significance/confidence-bound reference lines on the chart.
• First-order differencing of the series, with re-analysis of the differenced data.
• MA(1) series generation from user-supplied φ₁, variance, and sample size.
• MA(q) parameter estimation, using both Method of Moments and MSE/CSS minimization.
• AI feedback (Groq / Llama 3.3) that interprets the ACF and trend results in Persian.
• Dark/light mode, multi-page layout with routing.

Tech Stack

• Frontend: React 18 + TypeScript, Vite, React Router, shadcn/ui (Radix primitives) + Tailwind, Recharts for charts, xlsx for spreadsheet parsing.
• Backend: Python, Flask + Flask-CORS, NumPy/SciPy (parameter optimization), statsmodels (PACF), Groq API (AI feedback).

Project Structure

src/
  App.tsx                    # single-page analysis flow (upload -> analyze -> difference)
  pages/
    TimeSeriesPage.tsx       # analysis page (routed variant of App.tsx flow)
    MA1Generation.tsx        # MA(1) series generator page
    Layout.tsx                # shared layout/navbar shell
  components/
    Navbar.tsx
    FileUploadArea.tsx, AnalysisControls.tsx, ResultsDisplay.tsx, TimeSeriesChart.tsx
    ui/                       # shadcn/ui component library
  utils/
    timeSeriesAnalysis.py    # all analysis/estimation logic
    AI_engine.py              # Groq-based AI feedback
    types.ts
api_server.py                 # Flask API exposing the Python analysis functions

Setup

Frontend

npm i
npm run dev

Backend

pip install -r requirements.txt
python api_server.py       # runs on http://localhost:5000

AI feedback requires a GROQ_API_KEY in a .env file at the project root.

API Endpoints

Endpoint	Purpose
POST /api/parse-csv	Parse uploaded CSV content into values/labels
POST /api/analyze	Compute ACF, PACF, and trend coefficient
POST /api/difference	First-order difference the series and re-analyze it
POST /api/generate-ma	Generate a synthetic MA(1) series
POST /api/estimate-ma-parameters	Estimate MA(q) parameters (MME and MSE methods)
POST /api/ai-feedback	Get AI-generated Persian commentary on ACF/trend results
GET /api/health	Health check

Contributors

• ehssanhdp — Project scaffolding: React/Vite/TypeScript frontend setup, shadcn/ui component library, Flask API server, core analysis functions (ACF, trend coefficient), AI feedback engine (AI_engine.py), repo maintenance.
• Parsa Sarfarazi — MA(1) series generation feature, first-order differencing, multi-page routing/layout, styling overhaul.
• reyhane-yh — MA(q) parameter estimation (Method of Moments and MSE/CSS minimization) and its API endpoint.
• narimanvt — Partial autocorrelation (PACF) computation and its chart visualization, including significance/confidence-bound reference lines.