phone-usage-analysis

📱 Phone Usage Analysis (India)

Exploratory data analysis of 17,686 mobile users in India covering demographics, screen time, data usage, calls, app usage, spending and recharge cost, including a data-validation step that tests whether the data is realistic.

What It Does

• Loads and inspects the dataset (structure, missing values, duplicates, negative values)
• Validates the data logically (brand vs OS, activity time vs total screen time)
• Explores distributions and relationships with Matplotlib, Seaborn and Plotly
• Compares screen time across age groups, cities and primary use
• Tests correlations between all numeric variables

Dataset

phone_usage_india.csv: 17,686 users × 16 columns (age, gender, city, phone brand, OS, screen time, data usage, call duration, apps installed, social/streaming/gaming time, e-commerce spend, recharge cost, primary use). The notebook loads it automatically, so no upload is needed.

Key Findings

• Typical user: ~6.5 hrs screen time/day, 25 GB data/month, 151 min calls/day, 105 apps, ₹1,043 monthly recharge, ₹5,076 monthly e-commerce spend.
• No meaningful relationships: every pairwise correlation is near zero (strongest: −0.02).
• Groups look identical: average screen time differs by only ~0.1 hrs/day across age groups and ~0.25 hrs/day across cities.
• Balanced categories: each gender ≈ 33%, OS 50/50, each primary use ≈ 20%.
• Data quality issues: Apple phones appear on Android (879 of 1,775), and 76.5% of users have social + streaming + gaming time greater than their total screen time.

Conclusion: the dataset is structurally clean but shows the hallmarks of synthetic, randomly generated data, so it cannot support real-world claims about Indian mobile users. The project demonstrates a full EDA workflow and the habit of validating data before trusting it.

Tech Stack

Python · Pandas · NumPy · Matplotlib · Seaborn · Plotly · Google Colab

Run It

Click the Colab badge above (use File → Save a copy in Drive, then Runtime → Run all), or locally:

pip install pandas numpy matplotlib seaborn plotly jupyter
jupyter notebook phone_usage_analysis.ipynb

Future Improvements

• Repeat the analysis on real telecom or survey data
• Add statistical tests (e.g. ANOVA) across groups
• Build a Streamlit dashboard with filters