uncg-cse / volag Goto Github PK

70% of the data should be used for training.

After training various models, test them with test data.

Acquire appropriate weather data.

Train Random Forest model on data set.

Use sklearn.

Use TensorFlow to see how deep learning performs on the dataset.

12 month heat map visualization of flight delays as the year progress through various seasons. This would highlight the weather/seasonal impact on flight delays in various regions of the US.

Train a Decision Tree model on the data set.

Use sklearn

Train a Naive Bayes based model on the dataset.

Use GaussianNB from sklearn.

Use python (selenium) to scrape http://newa.cornell.edu/index.php?page=hourly-weather for hourly weather.

Determining which weather stations we need data for.

The datasets are pretty large. So, they should not be committed to the repository.

(for visibility)
@hahnalex @JoshCMoore @tlcox3 @somyamohanty

The blank fields are processed in as NaN by pandas. When attempting to sum them into a dictionary, the result of the sum is nan. I need to remedy this.

Train an SVM based model on the dataset.

Use sklearn.

Produce a correlation analysis and a heatmap of the flight data itself. For comparison to a correlation and heat map on the merged dataset.

The decided features for the models should be used to form a new labelled dataset.

70% of the data should be used for training the models.
30% of the data should be used for testing the models.

Connect the weather data to the flight data.