For instance, the model is trained with A,B,C,D,E features to predict F target. Then, I want to make prediction on a new data, but with only A,C,D,E is known without doing missing value imputation on B feature.
k-nearest neighbors algorithm (k-NN)- A model to be created for each input variable that has missing values. Although any one among a range of different models can be used to predict the missing values, the k-nearest neighbor (KNN) algorithm has proven to be generally effective, often referred to as “nearest neighbor imputation - The same method followed by all famous On line shopping websites and Google as well and will provide search out put /what you supposed to do /guess based on time frames/intervals - ex .Netflix /Amazon prime/Amazon shopping/google .com-
For this I use data mining models in SPSS. It will even allow multiple missing values. Not just a NaN option, but also different types of missing reasons (not answered, skipped).
XGBoost can be applied for that purpose. XGBoost is a state-of-the-art ML approach that is flexible during customization.
Hello Prasojo !! You can use SVM, but your results will not be accurate. In addition, you must have a machine with a good processing power and memory.
Thank you all for the answers. I tried to use Random Forest in Orange to do such, and it seems to be able to predict a class with incomplete features. I once read that RF can handle missing data by using average value or inference.
In the documentation here (https://orange3.readthedocs.io/projects/orange-data-mining-library/en/latest/reference/classification.html) it says:
"The inducer can handle missing values of attributes and target. For discrete attributes with more than two possible values, each value can get a separate branch (binarize=False), or values can be grouped into two groups (binarize=True, default)"
But I don't know what does that mean. My features are discrete values with different ranges, for example: one is in the range of 0-5000, another is in the range of 0-10, and so on.
Hope am not to late you can also check out catboost algorithm still part of random forest family. It can easily integrate with deep learning frameworks like Google's TensorFlow and Apple's Core ML. It has a wide range of function which covers the problem you stated.
Here is a link to the documentation https://catboost.ai/docs.
- Badges
- Science topic