I am a programmer not much into math.

I have daily data of multiple products (total sales count and amount).

Manually we can see the anomaly in the count or amount like if there is a dip or hike in data by plotting them.

But now we have lots of products and can not track them manually. So now I am looking for a solution that can detect anomalies in each product.

Like in the below picture, the blue line shows the daily count of one product for the last 30 days.

I also tried to calculate the Exponentially weighted Average (orange line and added margin in gray line).

Blue line shows the daily trend of data

Then I get the data for the same product for the same weekdays. like if data is of if will get the data of Mondays.

then took the same exponentially weighted averages.

Weekday's stats of product

Then I created the 3rd graph by adding the weightages of both graphs. like took 30% from first value and 70% from the second.

combined graphs

But still, I am not sure if it will work fine.

Or if there are better ways to detect anomalies.

I hope if I was able to describe the actual problem.

Please try to be kind as I am very new to this and don't have much knowledge. Some useful articles will also be helpful thanks.

  • 2
    $\begingroup$ Why closing proposals for this well written question ? $\endgroup$ Commented Jan 3, 2022 at 13:03
  • $\begingroup$ @JeanMarie because it lacks all the necessary details to actually provide any help. We know, from the question, nothing about the data, what it will be used for, who will be reviewing the summaries, etc. This cannot be answered as it stands without making an enormous number of guesses and assumptions. $\endgroup$
    – postmortes
    Commented Jan 5, 2022 at 7:03
  • $\begingroup$ Despite there are similar posts they had this issue with ARIMA model here & here as well as this old one concerning logged data I'm interested this question. $\endgroup$
    – Mario
    Commented Jan 6, 2022 at 0:33


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.