0
$\begingroup$

I am currently working on the COVID data of a US state to predict the number of people affected per day and also the number of patients is hospitalized per day by developing the Fuzzy Mamdani Model using Matlab. Our team gathered the official data provided by the Department of Health of the state. We have taken into account the data from July 1st to November 30th (total 153 days).

Among the variables considered in the model, the most missing values are the number of(PCR positive patients and Antigen positive patients). Data of around 45 days are missing of these 2 variables of the total of 153 days. The good thing is the summation of PCR positive and Antigen positive patients are the total number of affected patients which is available for 153 days.

Now to fill the missing values of 45 days we are considering the average values of the Total number of affected, number of PCR positive, and the number of (153-45)=108 days and calculating the percentage ratio. After that fill up the missing 45 days data of PCR and Antigen positive based on this ratio.

Our team is expecting your criticism on the proposed method and at the same time your proposal for alternative method to fill the missing values. Thanks for your cooperation.

$\endgroup$
0
$\begingroup$

regarding your method: It may work, but it feels that just averaging the 3 values for the 108 days, can mess up the knowledge you have in your data. Say you have some clear trend for 108 days with full data, using average you will lose the knowledge about these trend.

Maybe extract the ratio for every day of the 108 days, and then treat it as time series data. try predicting the ratio, which is reasonable in such cases, because the number of positives may depend on the number of total tests, where the last my depend on the day of the week. so try to predict the ration and then calculating the absolute numbers.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.