if __name__ == '__main__':
print("Reading file ...")
X_df, y_df = read_data(filename)
#no training, use all sample for test:
skf = ShuffleSplit(len(y_df), n_iter=1, test_size=0.99, random_state=57)
print("Training file ...")
for train_is, test_is in skf:
print '--------------------------'
# use dummy clustering
#tracker = Tracking.HitToTrackAssignment()
tracker = Clustering.ClusterDBSCAN(eps=0.5, rscale=0.001)
X_train_df = X_df.iloc[train_is].copy()
y_train_df = y_df.iloc[train_is].copy()
X_test_df = X_df.iloc[test_is].copy()
y_test_df = y_df.iloc[test_is].copy()
# Temporarily bypass splitting (need to avoid shuffling events)
X_test_df = X_df.copy()
y_test_df = y_df.copy()
tracker.fit(X_train_df.values, y_train_df.values)
y_predicted = tracker.predict(X_test_df.values)
# Score the result
total_score = 0.
'pt': [0.],
'phi': [0.],
'xVtx': [0.],
'yVtx': [0.]
})
data_track = data_track.drop(data_track.index[[0]])
df = pd.read_csv("hits_" + str(N) + ".csv")
y_df = df[['particle']]
# X_df = df.drop(['hit','layer','particle','event'], axis=1)
X_df = df.drop(['particle', 'layer', 'iphi'], axis=1)
#replace particle with 100000000*event+particle
y_train = df['particle'].values + df['event'].values * 1000
tracker = trk.ClusterDBSCAN(eps=0.004, rscale=0.001)
tracker.fit(X_df.values[:1000], y_train[:1000])
df_result = df.drop(['particle'], axis=1)
y_predicted = tracker.predict(df_result.values)
df_result = pd.concat(
[df_result, pd.DataFrame({'track': y_predicted})], axis=1)
for col in ['event', 'track']:
df_result[col] = df_result[col].astype('int32')
df_result_truth = pd.concat(
[df_result, pd.DataFrame({'particle': y_train})], axis=1)
df_result_truth['particle'] = df_result_truth['particle'].astype('int32')
df_result.to_csv("result_" + str(N) + ".csv",