Airlines Delays Analysis

Airlines Delays Analysis

Python data science project I came up with as a capstone to finish off Jose Portilla's course Python for Data Science and Machine Learning Bootcamp on Udemy. This is a portfolio project to showcase what I learned from the course.

Project Details

• Details
• Link to Demo
• Tools Used
• What I learned

Details

This is a five part analysis and series of machine learning models that dig into the Airlines Delays dataset from Kaggle. Loads data in, cleans as necessary for NaN or null values, and has a focused scope for each notebook.

Link to Demo

Links to each part of the analysis:

• Part 1: Exploration
• Part 2: Linear and Logistic Machine Learning
• Part 3: Decision Trees and Random Forests Machine Learning
• Part 4: K-Nearest Neighbors and K-Means Clustering Machine Learning
• Part 5: Principal Component Analysis

Tools Used

• Python
• Excel
• numpy
• pandas
• plotly
• matplotlib
• seaborn
• sklearn
• gcmap
• holoviews

What I learned

Below are some code snippets I'm proud of from this project:

Part 1: Exploration. Global map of flights in the data.

fig = go.Figure()

## Loop thorugh each flight entry to add line between source and destination
for slat,dlat, slon, dlon, num_flights in source_to_dest:
    fig.add_trace(go.Scattergeo(
                        lat = [slat,dlat],
                        lon = [slon, dlon],
                        mode = 'lines',
                        line = dict(width = num_flights/100, color="red")
                        ))


## Loop thorugh each flight entry to plot source and destination as points.
fig.add_trace(
    go.Scattergeo(
                lon = routes["long_from"].values.tolist() + routes["long_to"].values.tolist(),
                lat = routes["lat_from"].values.tolist() + routes["lat_to"].values.tolist(),
                hoverinfo = 'text',
                text = scatter_hover_data,
                mode = 'markers',
                marker = dict(size = 10, color = 'blue', opacity=0.1,))
    )

## Update graph layout to improve graph styling.
fig.update_layout(title_text="Connection Map Depicting Flights from Brazil to All Other Countries (Orthographic Projection)",
                  height=500, width=500,
                  margin={"t":0,"b":0,"l":0, "r":0, "pad":0},
                  showlegend=False,
                  geo= dict(projection_type = 'orthographic', showland = True, landcolor = 'lightgrey', countrycolor = 'grey'))

fig.show()

Part 2: Linear and Logistic Machine Learning. Linear regression model evaluation.

print('MAE:', metrics.mean_absolute_error(y_test, predictions))
print('MSE:', metrics.mean_squared_error(y_test, predictions))
print('RMSE:', np.sqrt(metrics.mean_squared_error(y_test, predictions)))

Part 3: Decision Trees and Random Forests Machine Learning. Classification report for Decision Trees model.

predictions = dtree.predict(x_test)

print(classification_report(y_test, predictions))

Part 4: K-Nearest Neighbors and K-Means Clustering Machine Learning Plotting the error rate vs tested K values.

plt.figure(figsize=(12,6))
plt.plot(range(1,41), error_rate, linestyle='--', marker='o', markerfacecolor='red', markersize=10)
plt.title('Error Rate vs K Value')
plt.xlabel('K')
plt.ylabel('Error Rate')

Part 5: Principal Component Analysis Plotting the different components found.

plt.figure(figsize=(8,6))
plt.scatter(x_pca[:,0], x_pca[:,1], c=df['Class'], cmap='plasma')
plt.xlabel('First principal component')
plt.ylabel('Second Principal Component')