3.6

// road map
https://data.vicroads.vic.gov.au/arcgis/rest/services/HeavyVehicles/OSOM_SCHEME_MAP_NETWORK_D6/FeatureServer

// signal
https://services2.arcgis.com/18ajPSI0b3ppsmMt/arcgis/rest/services/traffic_lights/FeatureServer

• average_precision_score：AP that calculates the predicted value
• f1_score: Calculate the F1 value, also known as the balanced F-score or F-meature
• precision_recall_curve：Precision-recall pairs that calculate different probability thresholds
• precision_score： calculates precision
• recall_score： calculates recall
• roc_auc_score:

• Data from gov of Victoria
• get weather from http://api.weather.com to get history weather data

• Feature includes accidents location, weather condition, road infrastructure, car and driver information. when accident happen, y = 1
• Simulate the datasets for the situations that accidents not happen in all conditions, y = 0

• Using Tensorflow to train, the result isn't good enough
• Using XGboost, the result better.