Timeseries Model
Here's an example output from the API. The last training data date is 11-19 so we predict up to 7 days (configurable) forward.
The mean predictions are the average predictions
The quantile_1 array is the low end of the prediction, all the way up to quantile_9 which represents a 90% confidence the
if you want to update the *data* but not the model itself, you can dynamically add data to the training data set (make sure it's annotated and verified). The new predictions will automatically move forward in time as new data is added. This will not retrain the model, it will just incorporate new data into the prediction (i.e. predicting a new day in the future if you add a new day to the training data set).
If you want to improve model performance, you can retrain the model with the additional data.
{
"preds": {
"forecaset_start": "2021-11-20 00:00:00",
"freq": "D",
"mean_predictions": [
327.8624572753906,
272.3104553222656,
273.4146423339844,
326.5408630371094,
316.8637390136719,
339.1827697753906,
329.3565979003906
],
"quantile_1": [
302.957763671875,
237.03968811035156,
242.19688415527344,
263.7484130859375,
276.9398193359375,
308.104248046875,
294.96807861328125
],
"quantile_5": [
325.3675842285156,
272.91619873046875,
269.1368713378906,
318.96649169921875,
327.41021728515625,
342.30682373046875,
330.6439208984375
],
"quantile_9": [
349.7281494140625,
301.29095458984375,
297.5340270996094,
359.11358642578125,
377.1477355957031,
374.61785888671875,
367.8771667480469
]
},
"success": true
}
Copy link