andrewwuan/PredictionIO-Churn-Prediction-H2O-Sparkling-Water

template result

Closed this issue · 3 comments

phiripatrick663 commented

Hi Andrew,
Happy to report that the new version template ran without errors. I do need a little assistance though, I have no idea what to expect as a result of the trial query. At first I was not sure how to query it, but made the query into a python file, made it executable and ran. Is that the correct way? Can you give an example of expected result? Any further suggestions on use and customizing template?

Thanks.

andrewwuan commented

@phiripatrick663 If you look at README.md, putting the query in a python file and running it is indeed the correct way. An example query would be (copying from README.md):

engine_client.send_query({'intlPlan': True, 'voiceMailPlan': True, 'numVmailMsg': 41, 'totalDayMins': 173.1, 
    'totalDayCalls': 85, 'totalDayCharge': 29.43, 'totalEveMins': 203.9, 'totalEveCalls': 107, 
    'totalEveCharge': 17.33, 'totalNightMins': 122.2, 'totalNightCalls': 78, 'totalNightCharge': 14.02, 
    'totalIntlMins': 10.0, 'totalIntlCalls': 15, 'totalIntlCharge': 3.94, 'customerServiceCalls': 0})

You can configure the set of arguments you provide by modifying Engine.scala and Algorithm.scala. I don't have a specific expected result, but generally, if you have a query that contains what's "like" the imported data, the output percentage will be higher, and vice versa.

phiripatrick663 commented

Thanks Andrew. I will play around with customizing those two files and wee what I can post as success.

andrewwuan commented

👍