This document details a data analysis project focused on driver performance within a trucking company, AIDispatcher, Inc. The project aims to uncover relationships between driver performance metrics and financial outcomes.
- Identify correlations between fuel efficiency, miles driven, earnings, workdays, and performance scores.
- Use real operational data and statistical techniques to reveal patterns.
- Inform decision-making and improve operational efficiency and driver evaluation.
- Weekly raw data on driver performance metrics is collected, cleaned, and aggregated.
- Analysis is performed on an aggregated dataset.
- Python libraries like pandas, seaborn, numpy, and matplotlib are used.
- Various statistical methods are employed, including correlation matrices (Pearson, Kendall, Spearman), heatmaps, boxplots, histograms, and scatterplots.
- Outlier detection and removal techniques are applied.
- Strong positive correlations exist between:
- Total miles driven and total gross earnings.
- Average miles driven and average gross earnings.
- Average workdays and average miles driven.
- Average profit/loss and performance score.
- Performance scores are more strongly correlated with profitability than with revenue or miles driven.
- Different correlation methods reveal variations in the strength of relationships, with Spearman and Kendall being more robust to outliers and capturing non-linear associations.
- Scatterplots show a positive correlation between average miles and average profit/loss, and a negative correlation between average fuel efficiency and average profit/loss.
- The analysis suggests that while increasing mileage can lead to higher profits, other factors like fuel efficiency and maintenance costs must be considered.
- Outliers are identified and analyzed to understand deviations from the general trends.
- Optimize route efficiency instead of simply maximizing mileage.
- Investigate the causes of declining fuel efficiency during extended work periods.
- Revise performance metrics to incentivize sustainable operational practices.
- Explore the relationship between work patterns and both revenue and efficiency.
The project concludes that a multi-method approach to correlation analysis provides a comprehensive understanding of the data relationships. The findings confirm core business dynamics while highlighting nuances that might be missed by using only one correlation method.
https://medium.com/@serorl0v/ranking-and-optimization-in-analytics-practical-examples-b790f088e426