🌦️ ForecastIQ — Weather Prediction Tool

AI-Powered Weather Forecasting with WHO Air Quality Integration

ForecastIQ is an intelligent ML-powered weather prediction platform built with Streamlit. Enter 14 environmental parameters — including location, meteorology, and WHO air-quality metrics — to instantly predict rainfall, temperature, and weather categories. Download a personalized, styled PDF report with every prediction.

---

🚀 Live Demo

Try the app now: https://forecastiq-weather-prediction-tool-2025.streamlit.app/

Get instant weather predictions by entering 14 environmental parameters including location, meteorology, and WHO air-quality metrics. Download your personalized weather report as a PDF!

forecaste.project.mp4

---

✨ Key Highlights

• 🌧️ Rain Occurrence Classifier — 92.99% accuracy (ExtraTrees) for predicting rainfall events
• 🌡️ Temperature Classifier — 97.72% accuracy (XGBoost) for temperature category prediction
• 📊 14-Parameter Input System — Geography, meteorology & all 6 WHO air-quality indicators
• 📄 PDF Report Generation — Download a beautifully styled personalized weather report
• 🔬 5-Algorithm Comparison — Extra Trees, Random Forest, XGBoost, LightGBM, SVM benchmarked
• 🌍 WHO Air Quality Integration — CO, Ozone, NO2, SO2, PM2.5, PM10 as prediction features
• 📓 End-to-End Notebook — Full training pipeline with evaluation charts and model export
• ⚡ Streamlit UI — Interactive, responsive web interface with instant predictions
• 📁 Dual Output Format — HTML report with PDF download (WeasyPrint / ReportLab fallback)
• 🔒 Reproducible Pipeline — Joblib-serialized models for consistent, version-safe inference

---

📋 Table of Contents

• Overview
• Features
• Project Structure
• Data & Input Features
• Models & Tasks
• Model Accuracy Results
• Technologies Used
• Installation
• Run the App
• Train & Export Models
• Troubleshooting
• Contact

---

🌐 Overview

ForecastIQ predicts temperature and rainfall based on environmental and air-quality variables, studying how WHO air-quality parameters relate to climate outcomes. The repository contains:

• A Streamlit UI (app.py) to make predictions with pre-trained models and download a styled HTML/PDF report.
• A Jupyter notebook (final1.ipynb) that loads the dataset, explores features, trains multiple algorithms (Extra Trees, RandomForest, XGBoost, LightGBM, SVM), evaluates results, and saves the best models with Joblib.

---

🔧 Features

🌦️ Core Prediction Features

• Rainfall Amount Regression — Predicts precise rainfall in mm using log-transform pipeline
• Temperature Regression — Continuous temperature prediction in Celsius
• Rain Occurrence Classification — Binary prediction: Rain / No Rain
• Temperature Category Classification — Multi-class: Cold / Moderate / Hot
• 14-Feature Input Interface — All geographic, meteorological, and air-quality parameters

🌍 WHO Air Quality Integration

• Carbon Monoxide (CO) — Air quality feature for climate correlation
• Ozone (O3) — Atmospheric ozone as a weather predictor
• Nitrogen Dioxide (NO2) — Pollution-weather relationship modeling
• Sulphur Dioxide (SO2) — Industrial emission impact on weather
• PM2.5 & PM10 — Particulate matter as environmental indicators
• Cross-domain Analysis — Studies how air quality drives climate outcomes

📊 ML Pipeline Features

• Multi-Algorithm Benchmarking — 5 algorithms evaluated side-by-side
• Log Transform Preprocessing — log1p de-skewing for rainfall regression
• KBinsDiscretizer Binning — Quantile-based temperature categorization
• Confusion Matrix Visualization — Per-class evaluation charts in notebook
• MSE & Accuracy Reporting — Regression and classification metrics
• Joblib Model Serialization — Portable .pkl models for deployment

📄 Report & Export Features

• Styled HTML Report — Full-page personalized prediction summary
• PDF Download — WeasyPrint (primary) or ReportLab (fallback) generation
• HTML Fallback Download — Always-available report export option
• Decorative Report Footer — Optional character.png branding in PDF

🖥️ UI/UX Features

• Responsive Streamlit Design — Works on desktop and tablet browsers
• Instant Prediction Feedback — Real-time results on form submission
• Input Validation — All 14 fields required with non-zero enforcement
• Clean Numeric Input Interface — Organized by feature category
• Browser Compatibility — Chrome, Firefox, Opera, Brave support

⚙️ Development & Production Features

• Cross-platform Compatibility — Windows 11, macOS, Linux
• Virtual Environment Support — Isolated dependency management
• Notebook-to-App Workflow — Train in notebook, deploy via Streamlit
• SDLC Best Practices — Structured development and evaluation pipeline
• Scalable Architecture — Extendable with new features and datasets

---

📁 Project Structure

ForecastIQ/
├── app.py                      # Streamlit application (inputs, predictions, report/PDF)
├── final1.ipynb                # End-to-end training & evaluation notebook
├── requirements.txt            # Dependency list
├── README.md                   # This file
├── __pycache__/                # Bytecode (ignored by git)
│
├── et_reg_rain.pkl             # ExtraTrees regression model (log rainfall)      [generated]
├── et_reg_temp.pkl             # ExtraTrees regression model (temperature)       [generated]
├── et_cls_rain.pkl             # ExtraTrees classifier (rain occurrence)         [generated]
└── et_cls_temp.pkl             # ExtraTrees classifier (temperature category)    [generated]

---

📐 Data & Input Features

Notebook dataset: IndianWeatherRepository.xlsx

• Read via pandas.read_excel — requires openpyxl

Targets created in the notebook:

Target	Type	Description
log_precip_mm	Regression	log1p(precip_mm) — de-skewed rainfall
rain_occurred	Classification	(precip_mm > 0).astype(int) — binary
Temperature bins	Classification	KBinsDiscretizer(n_bins=3, strategy='quantile')

App input features (14 numeric fields — all required):

Category	Features
Geography	latitude, longitude
Meteorology	humidity, wind_kph, cloud, pressure_mb, uv_index, feels_like_celsius
WHO Air Quality	air_quality_Carbon_Monoxide, air_quality_Ozone, air_quality_Nitrogen_dioxide, air_quality_Sulphur_dioxide, air_quality_PM2.5, air_quality_PM10

---

🤖 Models & Tasks

The workflow supports four prediction tasks:

#	Task	Type	Target
1	Rainfall Amount	Regression	log_precip_mm converted back via expm1
2	Temperature Value	Regression	temperature_celsius
3	Rain Occurrence	Binary Classification	No Rain / Rain
4	Temperature Category	Multi-class Classification	Cold / Moderate / Hot

Algorithms evaluated:

• ⭐ Extra Trees — Final .pkl models deployed in the app
• 🌲 Random Forest
• 🚀 XGBoost
• 💡 LightGBM
• 🔵 SVM (SVR for regression / SVC for classification)

---

📈 Model Accuracy Results

Rain Occurrence Classification

Model	Accuracy
⭐ ExtraTrees	92.99%
LightGBM	92.38%
XGBoost	92.08%
Random Forest	91.85%
SVM	87.28%

Temperature Category Classification

Model	Accuracy
⭐ XGBoost	97.72%
LightGBM	97.64%
Random Forest	97.03%
ExtraTrees	96.19%
SVM	92.61%

---

🛠️ Technologies Used

Backend & ML

Technology	Purpose
Python 3.10+	Core language
scikit-learn	Extra Trees, Random Forest, SVM, preprocessing
XGBoost	Gradient boosting classifier & regressor
LightGBM	Fast gradient boosting
Joblib	Model serialization (.pkl export/load)
NumPy / Pandas	Data manipulation and feature engineering
openpyxl	Excel dataset reading

Frontend & Reporting

Technology	Purpose
Streamlit	Interactive web UI
WeasyPrint	Primary PDF generation
ReportLab	PDF fallback generator
Matplotlib / Seaborn	Evaluation charts in notebook
Pillow	Image handling for PDF report

---

⚙️ Installation (Windows / PowerShell)

1. Clone the repository

git clone https://github.com/varshinijayaprabhu/ForecastIQ-Weather-Prediction-Tool.git
cd "ForecastIQ  Weather Prediction Tool"

2. Create and activate a virtual environment

python -m venv .venv
.\.venv\Scripts\Activate.ps1

3. Install dependencies

pip install -r requirements.txt

4. PDF support (choose one)

• WeasyPrint — already in requirements.txt; ensure Cairo & Pango native libs are installed on Windows
• pdfkit + wkhtmltopdf — download the binary from wkhtmltopdf.org and ensure it is in PATH

---

▶️ Run the App

streamlit run app.py

In the UI:

1. Enter all 14 feature values (non-zero required)
2. Click Predict to see rainfall (mm), temperature (°C), rain occurrence, and temperature class
3. Scroll to view the full HTML report
4. Click Download PDF (or "Download HTML" as fallback)

Required model files in the repo root:

et_reg_rain.pkl    et_reg_temp.pkl    et_cls_rain.pkl    et_cls_temp.pkl

---

📓 Train & Export Models (Notebook)

Open final1.ipynb and run cells in order:

1. Load data — Read IndianWeatherRepository.xlsx, inspect, and clean
2. Create targets — log_precip_mm, rain_occurred, temperature bins
3. Split & train — ExtraTrees, RF, XGBoost, LightGBM, SVM
4. Evaluate — MSE for regression; accuracy & confusion matrices for classification
5. Export models:

import joblib
joblib.dump(et_reg_rain,  "et_reg_rain.pkl")
joblib.dump(et_reg_temp,  "et_reg_temp.pkl")
joblib.dump(et_cls_rain,  "et_cls_rain.pkl")
joblib.dump(et_cls_temp,  "et_cls_temp.pkl")

---

📬 Contact

Designed and developed by Varshini J

---

⭐ If you found this project useful, consider giving it a star on GitHub!