Topics Covered:
- GCP AI and ML services overview
- BigQuery for data analytics
- Vertex AI for machine learning
- Cloud Functions and Cloud Run
Code Example - Toronto Traffic ML Pipeline:
# GCP Vertex AI pipeline for Toronto traffic prediction
from google.cloud import aiplatform
from google.cloud import bigquery
import pandas as pd
from sklearn.ensemble import RandomForestRegressor
import joblib
# Initialize GCP clients for Toronto region
aiplatform.init(
project='toronto-smart-city-ml',
location='northamerica-northeast1' # Montreal region (closest to Toronto)
)
class TorontoTrafficPredictor:
def __init__(self):
self.bq_client = bigquery.Client()
self.model = None
def extract_toronto_traffic_data(self):
"""
Extract Toronto traffic data from BigQuery
"""
query = """
SELECT
traffic_volume,
hour_of_day,
day_of_week,
weather_condition,
special_events,
construction_zones,
timestamp
FROM `toronto-smart-city-ml.traffic.historical_data`
WHERE location IN ('DVP', '401', 'Gardiner', 'QEW')
AND timestamp >= TIMESTAMP_SUB(CURRENT_TIMESTAMP(), INTERVAL 1 YEAR)
"""
df = self.bq_client.query(query).to_dataframe()
return df
def preprocess_data(self, df):
"""
Preprocess Toronto traffic data for ML model
"""
# Feature engineering for Toronto-specific patterns
df['is_rush_hour'] = ((df['hour_of_day'].between(7, 9)) |
(df['hour_of_day'].between(16, 18))).astype(int)
df['is_weekend'] = (df['day_of_week'].isin([0, 6])).astype(int)
# Toronto-specific events
df['is_leafs_game'] = 0 # Would be populated from events API
df['is_raptors_game'] = 0
df['is_tfc_game'] = 0
# Weather impact (Toronto climate considerations)
weather_mapping = {
'snow': 3, 'rain': 2, 'cloudy': 1, 'clear': 0
}
df['weather_impact'] = df['weather_condition'].map(weather_mapping)
return df
def train_model(self, df):
"""
Train traffic prediction model for Toronto
"""
features = [
'hour_of_day', 'day_of_week', 'weather_impact',
'construction_zones', 'is_rush_hour', 'is_weekend',
'is_leafs_game', 'is_raptors_game', 'is_tfc_game'
]
X = df[features]
y = df['traffic_volume']
# Train Random Forest model
self.model = RandomForestRegressor(
n_estimators=100,
max_depth=10,
random_state=42
)
self.model.fit(X, y)
# Save model to GCS
joblib.dump(self.model, 'toronto_traffic_model.pkl')
return self.model
def deploy_to_vertex_ai(self):
"""
Deploy trained model to Vertex AI
"""
# Create Vertex AI model
model = aiplatform.Model.upload(
display_name='toronto-traffic-predictor',
artifact_uri='gs://toronto-ml-models/traffic-model/',
serving_container_image_uri='gcr.io/cloud-aiplatform/prediction/sklearn-cpu.1-0:latest'
)
# Deploy to endpoint
endpoint = model.deploy(
machine_type='n1-standard-2',
min_replica_count=1,
max_replica_count=10
)
return endpoint
# Usage example
predictor = TorontoTrafficPredictor()
traffic_data = predictor.extract_toronto_traffic_data()
processed_data = predictor.preprocess_data(traffic_data)
model = predictor.train_model(processed_data)
endpoint = predictor.deploy_to_vertex_ai()
print("Toronto traffic prediction model deployed to Vertex AI")
print(f"Endpoint name: {endpoint.display_name}")
BigQuery Analytics for Toronto Data:
-- BigQuery analytics for Toronto smart city data
-- Analyzing TTC ridership patterns
WITH toronto_transit_stats AS (
SELECT
DATE(timestamp) as date,
EXTRACT(HOUR FROM timestamp) as hour,
route_id,
station_name,
SUM(ridership_count) as total_riders,
AVG(delay_minutes) as avg_delay
FROM `toronto-smart-city.transit.ridership_data`
WHERE DATE(timestamp) >= DATE_SUB(CURRENT_DATE(), INTERVAL 90 DAY)
GROUP BY 1, 2, 3, 4
),
rush_hour_analysis AS (
SELECT
route_id,
station_name,
CASE
WHEN hour BETWEEN 7 AND 9 THEN 'Morning Rush'
WHEN hour BETWEEN 17 AND 19 THEN 'Evening Rush'
ELSE 'Off-Peak'
END as time_period,
AVG(total_riders) as avg_ridership,
AVG(avg_delay) as avg_delay_minutes
FROM toronto_transit_stats
GROUP BY 1, 2, 3
)
SELECT
station_name,
time_period,
ROUND(avg_ridership, 0) as average_daily_riders,
ROUND(avg_delay_minutes, 2) as average_delay_minutes,
CASE
WHEN avg_delay_minutes > 5 THEN 'High Delay'
WHEN avg_delay_minutes > 2 THEN 'Moderate Delay'
ELSE 'On Time'
END as performance_category
FROM rush_hour_analysis
WHERE station_name IN (
'Bloor-Yonge Station',
'Union Station',
'King Station',
'St. George Station',
'Dundas West Station'
)
ORDER BY average_daily_riders DESC;
Hands-on Project:
Toronto AI-Powered Urban Planning - Build an end-to-end ML pipeline using GCP services to analyze Toronto urban data and provide insights for city planning decisions.
