I'm working on a project where I'm tasked to find anomalous data (count of people) across different dimensions (categorical i.e country, occupation and a few more) and different days.

Below is a sample of the data

count is count for people per day, country and occupation

How do I go about this? Any recommended Python libraries or models? I found lots of tutorials on multivariate time series analysis but my data isn't multivariate time series as the categorical variables in this dataset do not depend on time.

CodePudding user response：

You can try with LSTM, BiRNN, GRU with multivariable time-series prediction. You can use tensorflow or pytorch to build the model.

CodePudding user response：

Sklearn has multiple possibilities. You could take a look at Isolation Forest.