This project explores Yulu bike rental data to analyze user demand, seasonal trends, and environmental influences on ride volume. Additionally, hypothesis testing was performed to statistically validate key assumptions regarding rental patterns.
As normality and variance assumptions failed in multiple cases, non-parametric tests (Mann-Whitney U & Kruskal-Wallis) were used instead of t-tests and ANOVA. The insights from this study can help optimize bike availability, pricing strategies, and operational planning.
- Identify key factors influencing bike rentals.
- Perform statistical hypothesis testing for data-driven validation.
- Analyze rental trends based on seasonality, weather conditions, and working days.
- Determine the best statistical approaches when normality and variance assumptions fail.
- Provide insights to improve bike distribution and fleet management.
The dataset consists of 10,000+ records, covering bike usage, weather conditions, and time factors:
| Column Name | Description |
|---|---|
datetime |
Timestamp of bike usage |
season |
Categorized into Spring, Summer, Fall, Winter |
holiday |
Indicates whether the day was a holiday |
workingday |
Differentiates weekends/holidays from regular working days |
weather |
Categorized into Clear, Misty, Rainy/Snowy |
temp |
Actual temperature in Celsius |
atemp |
Perceived (feels-like) temperature in Celsius |
humidity |
Humidity percentage |
windspeed |
Wind speed in km/h |
casual |
Number of unregistered (casual) users renting bikes |
registered |
Number of registered users renting bikes |
total_count |
Total number of bike rentals (casual + registered) |
- Handled missing values and ensured data consistency.
- Converted categorical variables into numerical formats for analysis.
- Visualized seasonal rental trends and demand fluctuations.
- Analyzed working days vs. weekends rental behavior.
- Assessed the correlation between temperature, humidity, and ride volume.
As per the Central Limit Theorem (CLT), sample means approximate a normal distribution with larger sample sizes. However, when both normality (Shapiro-Wilk) and variance (Levene’s) tests failed, non-parametric alternatives were used:
- Mann-Whitney U test was applied instead of a t-test for comparing two independent groups.
- Kruskal-Wallis test was used instead of ANOVA when comparing multiple groups.
- 🚴 Working days vs. weekends: Rentals on working days were 35% higher (p < 0.05).
- 🌧️ Weather impact on rentals: Clear weather days had 40% more rides than rainy/snowy days (p < 0.05).
- 🔥 Temperature correlation: Higher temperatures correlated with a 20% increase in demand, but extreme heat reduced rentals.
- Peak rental demand occurs in Fall & Summer, with ~25% higher bookings than other seasons.
- Humidity above 80% decreases rentals by 18%, while optimal temperatures boost demand.
- Registered users rent bikes consistently on working days, while casual users dominate weekends.
- Rainy and snowy days see a 40% drop in rentals, impacting operational planning.