Scribd
Upload
44 views5 pages

Machine Learning Pipeline: Model Name Dependent Variable

This document describes building and using machine learning pipelines and models to predict flight delays and on-time arrival using BigQuery machine learning. It creates a linear regression model to predict arrival delay based on airline, origin airport, destination airport, departure delay, taxi out time, and distance. It then uses this model to predict delays for flights from specific airports. It also creates a logistic regression model to predict on-time arrival and uses it to make predictions. It performs queries to analyze delays by origin airport.

Uploaded by

pp
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
44 views5 pages

Machine Learning Pipeline: Model Name Dependent Variable

This document describes building and using machine learning pipelines and models to predict flight delays and on-time arrival using BigQuery machine learning. It creates a linear regression model to predict arrival delay based on airline, origin airport, destination airport, departure delay, taxi out time, and distance. It then uses this model to predict delays for flights from specific airports. It also creates a logistic regression model to predict on-time arrival and uses it to make predictions. It performs queries to analyze delays by origin airport.

Uploaded by

pp
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 5

Machine Learning pipeline

#standardsql
CREATE OR REPLACE MODEL `flight.arrdelay` model name
OPTIONS
(model_type='linear_reg', input_label_cols=['ARRIVAL_DELAY']) AS dependent variable

SELECT
ARRIVAL_DELAY , Independent variable
AIRLINE,

ORIGIN_AIRPORT ,
DESTINATION_AIRPORT ,
DEPARTURE_DELAY ,
TAXI_OUT ,
DISTANCE
FROM
`big-query-291822.flight.flights`
WHERE
ARRIVAL_DELAY IS NOT NULL
Machine Learning pipeline
• Use your Linear Regression Mod #standardsql
el to predict: SELECT * FROM ML.PREDICT(MODEL `flight.arrdelay`,
(
(1) all variables of the model, limit 10 SELECT
ARRIVAL_DELAY ,
(2) flights delayed when the origin is AIRLINE ,
ORD ORIGIN_AIRPORT ,
>> ‘ORD’ AS ORIGIN_AIRPORT DESTINATION_AIRPORT ,
DEPARTURE_DELAY ,
(3) flights delayed from Dallas Fortwo TAXI_OUT ,
rth to Los Angeles DISTANCE
FROM
>> ‘DFW’ AS ORIGIN_AIRPORT `flight.flights`
WHERE
‘LAX’ AS DESTINATION_AIRPORT ARRIVAL_DELAY IS NOT NULL
LIMIT 10))
Create a logistic regression model to predict if the flight is on time or
is delayed based on carrier, origin, dest, dep_delay, taxi_out, distan
ce

建模
#standardsql
CREATE OR REPLACE MODEL `flight.ontime`
OPTIONS
(model_type='logistic_reg', input_label_cols=['ON_TIME']) AS
SELECT
IF(ARRIVAL_DELAY<= 15, 1 , 0) AS ON_TIME,
AIRLINE,
ORIGIN_AIRPORT ,
DESTINATION_AIRPORT ,
DEPARTURE_DELAY ,
TAXI_OUT ,
DISTANCE
FROM
`big-query-291822.flight.flights`
WHERE
ARRIVAL_DELAY IS NOT NULL
预测
• Use your Linear Regression SELECT * FROM ML.PREDICT(MODEL
Model to predict: `flight.ontime`,
(1) all variables of the model, limit 10 (
SELECT
(2) flights delayed when the origin is ARRIVAL_DELAY ,
ORD AIRLINE ,
>> ‘ORD’ AS ORIGIN_AIRPORT ORIGIN_AIRPORT ,
(3) flights delayed from Dallas DESTINATION_AIRPORT ,
Fortworth to Los Angeles DEPARTURE_DELAY ,
>> ‘DFW’ AS ORIGIN_AIRPORT TAXI_OUT ,
DISTANCE
‘LAX’ AS DESTINATION_AIRPORT FROM
`flight.flights`
WHERE
ARRIVAL_DELAY IS NOT NULL
LIMIT 10))
SELECT SELECT
ORIGIN_AIRPORT, COUNT(DEPARTURE_DELAY)AS totaldelay
COUNT(*) AS totaldelay FROM `big-query-291822.flight.flights`
FROM `big-query-291822.flight.flights` WHERE ORIGIN_AIRPORT = "ORD" AND MONTH = 7
WHERE ORIGIN_AIRPORT = "ORD" AND MONTH = 7 AND
DEPARTURE_DELAY > 0
GROUP BY ORIGIN_AIRPORT

SELECT
ORIGIN_AIRPORT,
AVG(DEPARTURE_DELAY) AS avg_dep_delay,
AVG(ARRIVAL_DELAY) AS avg_arr_delay
FROM `big-query-291822.flight.flights`
GROUP BY
ORIGIN_AIRPORT
ORDER BY
avg_arr_delay

You might also like