Ryanair Internship

# import packages import pandas as pd import matplotlib.pyplot as plt import seaborn as sns import plotly.graph_objects as go import numpy as np

# load data df = pd.read_csv('training.csv', sep=r'\t')

# view data df.head()

df.info()

df.index = df['FlightNumber'] df['DepartureDate'] = pd.to_datetime(df['DepartureDate']) df.drop(columns=['FlightNumber', 'DepartureYear', 'DepartureMonth', 'DepartureDay'], inplace=True) df.replace('(null)', np.nan, inplace=True) df['DepartureAirport'] = df['DepartureAirport'].astype('category') df['ArrivalAirport'] = df['ArrivalAirport'].astype('category') df['Route'] = df['Route'].astype('category')

df.isna().sum() / len(df)

weight_mean = int(df.loc[~df['ActualTOW'].isna(), ['ActualTOW']].astype(int).mean()) df.replace({'ActualTOW': np.nan}, weight_mean, inplace=True)

df['ActualTOW'] = df['ActualTOW'].astype(int)

# copy data df_test = df.copy()

# clean the data fully df_test.dropna(inplace=True)

# change types df_test['FLownPassengers'] = df_test['FLownPassengers'].astype(int) df_test['BagsCount'] = df_test['BagsCount'].astype(int) df_test['FlightBagsWeight'] = df_test['FlightBagsWeight'].astype(int)

# view new data df_test.info()

sns.pairplot(df_test)

sns.heatmap(df_test.corr())

df.drop(columns=['BagsCount', 'FlightBagsWeight'], inplace=True)

passengers_median = int(df.loc[~df['FLownPassengers'].isna(), ['FLownPassengers']].astype(int).median()) df.replace({'FLownPassengers': np.nan}, passengers_median, inplace=True)

df['FLownPassengers'] = df['FLownPassengers'].astype(int)

df['ActualTotalFuel'] = np.log(df['ActualTotalFuel'])

df.info()

len(pd.unique(df['DepartureAirport']))

len(pd.unique(df['ArrivalAirport']))

df.drop(columns=['ArrivalAirport', 'DepartureAirport', 'Route'], inplace=True)

df.info()

df['weekday'] = df['DepartureDate'].map(lambda x: x.weekday())

sns.catplot(data=df, x='weekday', y='ActualTOW', kind='box')

df.drop(columns=['weekday', 'DepartureDate'], inplace=True)

from sklearn.model_selection import cross_val_score, GridSearchCV from sklearn.linear_model import ElasticNet

grids = GridSearchCV(ElasticNet(), param_grid={'alpha': np.linspace(0, 1, 6), 'l1_ratio': np.linspace(0, 1, 11)}) grids.fit(df.loc[:, df.columns != 'ActualTOW'], df['ActualTOW'])

grids.best_estimator_

from sklearn.linear_model import LinearRegression

model = LinearRegression() cross_val_score(model, df.loc[:, df.columns != 'ActualTOW'], df['ActualTOW'])

from sklearn.ensemble import RandomForestRegressor

grids_forest = GridSearchCV(RandomForestRegressor(), param_grid={'n_estimators': np.linspace(100, 500, 5, dtype='int'), 'max_depth': np.linspace(2, 10, 5, dtype='int')}) grids_forest.fit(df.loc[:, df.columns != 'ActualTOW'], df['ActualTOW'])

grids_forest.best_estimator_

model = RandomForestRegressor(n_estimators=400, max_depth=10) cross_val_score(model, df.loc[:, df.columns != 'ActualTOW'], df['ActualTOW'])

best_model = RandomForestRegressor(n_estimators=400, max_depth=10) best_model.fit(df.loc[:, df.columns != 'ActualTOW'], df['ActualTOW'])

test = pd.read_csv('validation.csv', sep='\t')

test.info()

test = test[['ActualFlightTime', 'ActualTotalFuel', 'FLownPassengers']]

test.replace('(null)', np.nan, inplace=True)

test.isna().sum()

passengers_median = int(test.loc[~test['FLownPassengers'].isna(), ['FLownPassengers']].astype(int).median()) test.replace({'FLownPassengers': np.nan}, passengers_median, inplace=True) test['ActualTotalFuel'] = np.log(test['ActualTotalFuel'])

pred = best_model.predict(test)

pd.concat([pd.read_csv('validation.csv', sep='\t'), pd.DataFrame({'ActualTOW': pred})], axis=1).to_csv('validation.csv', sep='\t')