Baebon-Advisor - Suggested Timeframe bug

[TomasKopunec]

For the given testing scenario:

Find best time for 15/01/2024 in eastus between 12:00 and 18:00 (no sampling);

Given output of the carbon-sdk API:

[
    {
        "location": "eastus",
        "time": "2024-01-14T23:30:00+00:00",
        "rating": -2,
        "duration": "04:00:00"
    },
    {
        "location": "eastus",
        "time": "2024-01-15T03:30:00+00:00",
        "rating": 2,
        "duration": "04:00:00"
    },
    {
        "location": "eastus",
        "time": "2024-01-15T07:30:00+00:00",
        "rating": 3,
        "duration": "04:00:00"
    },
    {
        "location": "eastus",
        "time": "2024-01-15T11:30:00+00:00",
        "rating": 1,
        "duration": "04:00:00"
    },
    {
        "location": "eastus",
        "time": "2024-01-15T15:30:00+00:00",
        "rating": 6,
        "duration": "04:00:00"
    },
    {
        "location": "eastus",
        "time": "2024-01-15T19:30:00+00:00",
        "rating": 5,
        "duration": "04:00:00"
    },    
    {
        "location": "eastus",
        "time": "2024-01-15T23:30:00+00:00",
        "rating": 4,
        "duration": "04:00:00"
    },
    {
        "location": "eastus",
        "time": "2024-01-16T00:30:00+00:00",
        "rating": -3,
        "duration": "04:00:00"
    }
]

Note: Notice the surrounding data for 14/01 and 16/01 around the relevant data range

Expected: Suggest the time of 11:30 with score 1 on 15/01 Actual: Suggest the time of 00:30 with score -3 on 16/01

[jimbou]

I am a bit confused . is the error: 1)the fact that the returned datetime is not 15/01/2024 between 12:00 and 18:00 ? 2) the fact that rating is negative?

[jimbou]

The issue was with a bug in the testing not bug with the model.