| 🎯 Goal | Predict app ratings, classify review sentiment, and forecast category growth |
| 🧠 Approach | Supervised + unsupervised ML, NLP sentiment analysis, time-series forecasting |
| 📊 Data | Google Play Store apps dataset + user reviews dataset (Kaggle) |
| 📈 Delivery | 5 models spanning classification, regression, clustering & ARIMA forecasting |
App publishers live and die by ratings and reviews. This project answers:
⭐ What features actually drive a higher app rating? 💬 What is the sentiment mix hidden in user reviews? 💰 Do paid apps really outperform free ones? 🔮 Which app categories will grow over the next 5 years?
Answering these guides product decisions, pricing, and portfolio strategy.
| Dataset | Contents | Key Fields |
|---|---|---|
| 📱 Apps | Google Play app catalogue (Kaggle) | App, Category, Rating, Reviews, Installs, Type, Price, Content Rating, Genres |
| 💬 User Reviews | Review text + sentiment labels | Translated Review, Sentiment, Polarity, Subjectivity |
Preprocessing: missing-value handling, categorical encoding, numeric conversion, and merging the two datasets.
PREDICTIVE ANALYTICS PIPELINE (Python · scikit-learn · NLTK · statsmodels)
──────────────────────────────────────────────────────────────────────────
1. Clean & merge → handle nulls, encode categoricals, join apps + reviews
2. Classify sentiment → Random Forest Classifier (positive / neutral / negative)
3. Cluster apps → K-Means on rating tiers
4. Predict ratings → Random Forest Regressor (feature importance)
5. Free vs Paid → Logistic Regression
6. Forecast growth → ARIMA (5-year category projection)
Review sentiment mix, rating distribution and the top drivers of app ratings, built from the project's ML analysis.
- Positive reviews dominate the sentiment mix, with negative reviews outnumbering neutral ones.
- Most apps cluster in the 4.0–4.7 rating band — very few fall below 3.0 (K-Means tiers).
- Number of reviews is the strongest driver of an app's rating; installs and genres are moderate; price and content rating are minimal.
- Strong correlations surfaced: Price ↔ Type 0.78 and Genre ↔ Category 0.70, while rating shows only weak single-feature correlation.
- Free apps frequently match or beat paid apps on sentiment — user experience outweighs price; paid apps show marginally higher, less-variable ratings.
- ARIMA forecasts highlight emerging high-growth categories over a 5-year horizon.
| Stakeholder | Recommendation |
|---|---|
| 🛠️ Product Teams | Invest in UX & review volume — the real rating drivers, not price |
| 💰 Pricing | "Free + great experience" can out-perform paid; test before charging |
| 📈 Strategy | Use ARIMA category forecasts to prioritise the next portfolio bets |
| 💬 Support | Monitor negative-review share as an early churn / quality signal |
| Category | Tools |
|---|---|
| Language | Python |
| ML | scikit-learn (Random Forest, Logistic Regression, K-Means) |
| NLP | NLTK (sentiment analysis) |
| Forecasting | statsmodels (ARIMA) |
| Data & Viz | Pandas, NumPy, Matplotlib, Seaborn |
Google Play Store Predictive Analytics/
├── 📁 assets/
│ ├── 🎨 banner.svg # Repository banner
│ └── 📊 dashboard.svg # Reviews & ratings dashboard
├── 📁 docs/
│ └── 📄 Final Document.pptx # Project presentation & results
└── 📝 README.md # Project overview
Heta Chavda — Data Analytics | Machine Learning | Business Intelligence
⭐ Found this useful? Give it a star!