Attribute-Associated Neuron Modeling and Missing Value Imputation for Incomplete Data

Algorithm 1

The imputation based on ACFM and UMVDT.

INPUT: complete dataset , missing rate, ACFM, learning rate , maximum rounds T.

OUTPUT: the imputation error of at specified missing rate.

Generate an incomplete dataset according to specified missing rate.

Initialize missing values as variables, model weights, and thresholds.

Set t =0, precision =1.

while t<T and precision<0.001 do.

 for x in :

  Input x into model and get output y.

  Calculate the error for updating the model parameters and missing value variables respectively.

 end for

 Reconstruct model output and predict missing values.

 Calculate the imputation error and precision.

end while

Output the imputation error.