train_test_split

def train_test_ods(test_size=0.1):
    groups_volume=data.groupby(["origincode","destinationcode"])["passengers"].sum()
    result_df = groups_volume.reset_index().sort_values('passengers', ascending=False)
    from sklearn.model_selection import train_test_split
    train_df, test_df = train_test_split(result_df, test_size=test_size, random_state=42)

    train_org=train_df.origincode.values
    train_dst=train_df.destinationcode.values

    test_org=test_df.origincode.values
    test_dst=test_df.destinationcode.values

    import pandas as pd
    train_data=[]
    for org,dst in zip(train_org,train_dst):
        train_data.append(data[(data.origincode==org) & (data.destinationcode==dst)])
    train_data=pd.concat(train_data)

    test_data=[]
    for org,dst in zip(test_org,test_dst):
        test_data.append(data[(data.origincode==org) & (data.destinationcode==dst)])
    test_data=pd.concat(test_data)
    return train_data,test_data