Practice loading, cleaning, joining, and analyzing data with Python and pandas. Find free datasets, project ideas, and documented downloads.
Start with bike sharing for time-based grouping and charts. Explore ecommerce for joins, marketing for a documented cleaning challenge, or subscriptions for customer activity over time.
| Dataset | Preferred file | Practice focus | Time series |
|---|---|---|---|
| Ecommerce Sales & Returns | Parquet or source CSV | Merging customers, orders, items, payments, and refunds | Available |
| SaaS Subscriptions & Churn | Parquet or source CSV | Subscription periods, cohorts, invoices, and payment aggregation | Strong |
| Marketing Campaigns & Conversions | Parquet or source CSV | Category normalization, date parsing, deduplication, and attribution | Available |
| Inventory & Supply Chain | Parquet or source CSV | Movement ledgers, cumulative balances, and supplier comparisons | Strong |
| Bike Sharing Demand | Parquet or source CSV | Time-series grouping, feature engineering, and demand exploration | Strong |
Load the prepared Bike Sharing Parquet file, check its shape and coverage, then calculate an hourly profile.
import pandas as pd
df = pd.read_parquet("bike-sharing-v1.0.0-analysis.parquet")
df["dteday"] = pd.to_datetime(df["dteday"])
print(df.shape, df["dteday"].min(), df["dteday"].max())
hourly = df.groupby("hour", as_index=False).agg(
average_rentals=("total_rentals", "mean")
)
print(hourly.head())Load the selected file and inspect its columns, types, missing values, and sample records. Check the date range and identifiers before joining tables or calculating a result. Keep the original download unchanged so your preparation steps remain reproducible.
Start with the business question and dataset version. Show the loading and transformation steps, explain the important definitions, and finish with a clear result and its limitations.
No. Many datasets here are designed for analysis, reporting, and modeling practice. Synthetic business patterns should not be presented as evidence that a model will perform well on real-world records.
Work through Python exercises that help you inspect, transform, and explain data—not just load a file.