Data description

README.md

@@ -0,0 +1,3 @@
+
+![frequency of departures at given hour](./pics/departure_time_frequency.png)
+![popularity of trip destinations](./pics/stop_headsign_popularity.png)

prepare-ztm-data.sh

@@ -29,4 +29,8 @@ for k in "${keep[@]}"; do
 	fi
 done
 
+if [ ! -f "stop_times.normalized.tsv" ]; then
 	./normalize <stop_times.tsv >stop_times.normalized.tsv
+	./split_train_valid_test.py
+fi
+./stats.py

split_train_valid_test.py

@@ -1,12 +1,11 @@
+#!/usr/bin/env python3
 import pandas as pd
 from sklearn.model_selection import train_test_split
 
-TEST_SIZE = 25
-VALID_SIZE = 25
-
 data = pd.read_csv('./stop_times.normalized.tsv', sep='\t')
-train, test = train_test_split(data, test_size=TEST_SIZE+VALID_SIZE)
+
-valid, test = train_test_split(test, test_size=TEST_SIZE)
+train, test = train_test_split(data, test_size=0.5)
+valid, test = train_test_split(test, test_size=0.5)
 
 train.to_csv('stop_times.train.tsv', sep='\t')
 test.to_csv('stop_times.test.tsv', sep='\t')

stats.py

@@ -0,0 +1,52 @@
+#!/usr/bin/env python3
+import math
+import os
+import pandas as pd
+import contextlib
+
+pd.set_option('display.float_format', lambda x: '%.5f' % x)
+
+def float2time(d: float):
+    hours = math.floor(d * 24)
+    minutes = math.floor(d * 24 * 60 - hours * 60)
+    return "%s:%s" % tuple(str(x).rjust(2,'0') for x in (hours, minutes))
+
+data = pd.read_csv(f'./stop_times.normalized.tsv', sep='\t', dtype={ 'departure_time': float, 'stop_id': str, 'stop_headsign': str })
+
+print("--- Pictures -------------------------------------------------")
+
+with contextlib.suppress(Exception):
+    os.mkdir("pics")
+
+(data["departure_time"] * 24).plot(kind='hist', title="Częstotliwość czasu odjazdu").get_figure().savefig('pics/departure_time_frequency.png')
+print("pics/departure_time_frequency.png")
+data["stop_headsign"].value_counts().plot(kind='pie', title="Popularność celu").get_figure().savefig('pics/stop_headsign_popularity.png')
+print("pics/stop_headsign_popularity.png")
+
+
+print("--- Minmum departure time per stop headsign ------------------")
+shgroup = data.groupby('stop_headsign').min(numeric_only=True)
+shgroup["departure_time"] = shgroup["departure_time"].map(float2time)
+print(shgroup)
+print()
+
+print("--- Maximum departure time per stop headsign -----------------")
+shgroup = data.groupby('stop_headsign').max(numeric_only=True)
+shgroup["departure_time"] = shgroup["departure_time"].map(float2time)
+print(shgroup)
+print()
+
+print("--- Mean departure time per stop headsign --------------------")
+shgroup = data.groupby('stop_headsign').mean(numeric_only=True)
+shgroup["departure_time"] = shgroup["departure_time"].map(float2time)
+print(shgroup)
+print()
+
+
+print("--- Normalized data statistics -------------------------------")
+print(data.describe(include='all'))
+
+for subset in ['train', 'valid', 'test']:
+    print(f"--- {subset.title()} data statistics -------------------------------")
+    data = pd.read_csv(f'./stop_times.{subset}.tsv', sep='\t', dtype={ 'departure_time': float, 'stop_id': str, 'stop_headsign': str })
+    print(data.describe(include='all'))