Widget Predictions error about orange3 HOT 2 CLOSED

Ok, if I understand your issue, you are pointing out that Predictions show bad results if you train it on one k-means clustering and predict it on another k-means clustering.

Well, this is expected. The clusters you get may have different meanings (if nothing else, their names can be swapped). Thus, a model built on clustering won't perform well on another unless the clusterings are almost identical.

If I misunderstood your issue, please comment.

from orange3.

Ok, se eu entendo seu problema, você está apontando que as previsões mostram resultados ruins se você treiná-lo em um agrupamento k-means e prevê-lo em outro agrupamento k-means.

Bem, isso é esperado. Os clusters que você obtém podem ter significados diferentes (se nada mais, seus nomes podem ser trocados). Assim, um modelo construído em clustering não terá um bom desempenho em outro, a menos que os clusterings sejam quase idênticos.

Se eu entendi mal o seu problema, por favor, comente.

Hello friend, thanks for asking!

Another friend told me that my use case should not be treated with Kmeans, but with ARIMA, because according to him, my use case requires time series analysis!

I will attach the spreadsheets that I need to predict, to make it clearer and more objective!
AULAS_EXPANDIDAS.xlsx
AULAS_EXPANDIDAS_PREVISAO.xlsx

the result of line 35 is = CALVICIE 3, FACIAL 5, DESIGNER 6 and PEDICURE 8:
3, 5, 6 and 8!

Please, can you generate a project that makes this prediction based on the spreadsheets I am sending?

I've tried everything I can, but I can't!

from orange3.