108 lines (80 loc) · 4.09 KB

Aviation KPI Data Science Project

1. Machine Learning Phase (ML_Phase)

Purpose:

To develop a predictive model for analyzing key aviation performance metrics and optimizing operational efficiency.

Key Steps:

Feature Engineering:

Processed raw aviation KPI data to extract meaningful features (e.g., flight delays, turnaround time, fleet utilization).
Standardized and normalized numerical variables for improved model performance.

Model Selection:

Ridge Regression was chosen due to its ability to handle multicollinearity and prevent overfitting by imposing an L2 regularization penalty.
Why not Linear Regression?
- The dataset exhibited multicollinearity among features such as fuel efficiency, turnaround time, and aircraft utilization.
- Ridge Regression mitigates this issue by shrinking coefficients, leading to a more stable model.
Why not Logistic Regression?
- Logistic Regression is suitable for classification tasks, whereas this project required predicting continuous variables (e.g., revenue, cost, profit).

Hyperparameter Tuning:

GridSearchCV was employed to find the optimal alpha parameter for Ridge Regression, balancing bias and variance.

Evaluation Metrics:

Used Mean Squared Error (MSE) and R-squared to measure model accuracy and explanatory power.
Ridge Regression provided better generalization compared to a standard linear regression model.

Output:

A robust predictive model capable of estimating key aviation performance indicators such as revenue, operating costs, and net profit margins.

2. Data Cleaning Phase (Cleaning_Phase)

Purpose:

To ensure high-quality data by handling inconsistencies, outliers, and missing values.

Key Steps:

Outlier Detection:

Applied statistical methods (e.g., IQR, Z-score) to identify and remove anomalies in fuel efficiency, fleet utilization, and delay times.

Missing Data Handling:

Imputed missing values using:
- Mean/Median for numerical variables (e.g., turnaround time, aircraft utilization).
- Mode for categorical variables (e.g., airline, aircraft type).

Tools Used:

pandas, numpy

Output:

A clean dataset prepared for machine learning and SQL integration.

3. SQL Phase

Purpose:

To integrate MySQL for structured querying and analysis of aviation KPIs.

Key Steps:

Data Extraction:

Imported the cleaned dataset into MySQL for structured storage.

Data Aggregation:

Utilized GROUP BY, SUM, AVG to calculate:
- Average turnaround time per airline
- Total revenue and cost per flight
- Monthly profitability trends

Filters:

Applied WHERE and HAVING clauses to filter for high-revenue flights, significant delays, and fleet performance.

Database Creation:

Created a relational database to store flight details, revenue, costs, and delay statistics.

Output:

Well-structured MySQL tables ready for analysis and visualization.

4. Power BI Visualization Phase

Purpose:

To create interactive dashboards for data-driven decision-making.

Key Insights Extracted:

Total Profit: 1.03bn USD
Total Flights Analyzed: 200,000
Total Revenue: 5.11bn USD
Average Delay: 59.36 minutes
Fleet Utilization Rate: 12.00%

Key Visualizations:

Flight Delays Analysis: Breakdown of major, minor, and on-time flights.
Daily Trends: Delay variations by time of day.
Revenue and Cost Insights: Comparison of operating costs and revenue across different efficiency levels.
Financial Performance Breakdown: Profitability trends and revenue contribution by flight.
Monthly Profitability Trends: Visualization of revenue vs. cost per month.

Interactivity:

Added slicers and filters to explore data dynamically.

Output:

A Power BI dashboard that provides a comprehensive analysis of aviation KPIs.

General Notes:

The project workflow is modular, enabling iterative improvements.
Tools Used: Python (pandas, seaborn, sklearn), MySQL, Power BI
Key Focus: Delivering actionable insights aligned with business goals.