Apply model to data stream
Let's now apply your trained machine learning Pipeline to streaming data, and categorize the values immediately.
You'll then use predict()
on the incoming messages to determine the category.
Based on the result of the prediction you will take action, and close the windows in your house (or not).
Remember that category 1 means good weather, whereas category 0 signifies bad, cold weather.
Additionally, the pipeline returns an array of predictions. As you passed in only one element, you need to access the first element using category[0]
.
The function close_window()
will handle this for you, and will additionally log the record for further study.
pandas
as pd
and json
have been preloaded the session for you, and the model is available as pl
.
This exercise is part of the course
Analyzing IoT Data in Python
Exercise instructions
- Parse the dictionary into a pandas DataFrame with
DataFrame.from_records()
"timestamp"
as index, andcols
as columns. - Determine the category of this record by using
predict()
from the pipeline object and assign the result tocategory
. - Call
close_window()
with the DataFramedf
as the first argument, andcategory
as the 2nd argument.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
def model_subscribe(client, userdata, message):
data = json.loads(message.payload)
# Parse to DataFrame
df = pd.____.____([data], index=____, columns=____)
# Predict result
category = ____
if category[0] < 1:
# Call business logic
____
else:
print("Nice Weather, nothing to do.")
# Subscribe model_subscribe to MQTT Topic
subscribe.callback(model_subscribe, topic, hostname=MQTT_HOST)