admission probability based on Anomaly Detection

KNIME Analytics Platform
#1

HiExample_DB.xlsx (10.0 KB) ,
I have a university student database. Usually, each student has 3 or 4 school grades from 19 different school courses. Therefore each student has 19-3(4)=16(15) missing values. There are names of university study programs also. I need to predict for school pupils:

  1. admission probability for each study program
  2. how “typical” student he will be in each study program.
    School pupils usually have all grades.
    values ​​of each grade from 1 to 10. there is a positive correlation between school grades.
    I’m trying to use Anomaly Detection algorithm e.g. EXAMPLES/50_Applications/39_Fraud_Detection/Keras_Autoencoder_for_Fraud_Detection_Training
    But node Keras Network Learner doesn’t work with missing values.
    How to replace (restore) the missing value? or maybe there is another way to predict admission likelihood?
#2

These are very broad questions.

There are a wide variety of strategies for dealing with missing values. This is something that data scientists spend a LOT of time thinking about, and the “optimal” method depends on the dataset, the analysis, and the specific end application. Here’s an overview of some methods: 1, 2, 3.

There are also a wide variety of prediction methods, and there’s often no one “right” answer. You’ll have to do some reading on the advantages and disadvantages of each of these algorithms, and watch some of the KNIME videos on predictive modeling for tips on how to implement them.

1 Like