SMOTE | Synthetic Minority Over-sampling Technique | Machine Learning library

You need to instantiate ASSearch and ASEvaluation objects. If you have command-lines, you can use the from_commandline helper method like this:

You can use plt.bar for a bar plot. By drawing two bar plots onto the same subplot, the first still is partially visible.

Your pipeline has two fitted steps (+ the scaler): the SMOTE augmentation and the random forest. It looks like this is confusing the eli5 which wants to work with the assumptions that only the last layer is fitted. To get the weight explanation of the random forest you could try calling eli5 only on that layer of the pipeline with

It seems that you are plotting your heatmap with Seaborn. You can format numbers with seaborn.heatmap's fmt argument. Doing cm_plot = sns.heatmap(cm, annot=True, cmap='Blues', fmt='d') should work.

SMOTE requires the values in each categorical/numerical column to have uniform datatype. Essentially you can not have mixed datatypes in any of the column in this case your biz_category column. Also merely casting the column to categorical type does not necessarily mean that the values in that column will have uniform datatype.

One possible solution to this problem is to re-encode the values in those columns which have mixed data types for example you could use lableencoder but I think in your case simply changing the dtype to string would also work.

Below is an explanation of what could be the best approach for your case.

• Usually SMOTE balances out the data by random upsampling, so even if you have a data sample distribution like Class A having 15000 Records and Class B having 200 records it would upsample the Class B to 15000 Records too.
• Having too many random samples generated from the 200 Records it self sometimes makes the model very hard to learn and differentiate between classes, since the upsampling has significantly increased Class B records from 200 to 15000 by duplicating it.

1. Instead of SMOTE I would recommend to try Stratified Sampling between the train/test and then try building the model on top of it.
2. Having class weights as parameter is another best approach and its present almost for all ML algorithms. In your case for Keras you can Refer Here it could be very helpful.

If you stored the TPOTClassifier in the variable my_tpot, then you can access the final trained pipeline by accessing the fitted_pipeline_ attribute:

Your need to split your pipeline into 2 parts : one to process the numeric features (with the min max scaler) and another one to process categorical features (with the one hot encoder). You can use the class ColumnTransformer from scikit-learn : https://scikit-learn.org/stable/auto_examples/compose/plot_column_transformer_mixed_types.html

You are correct that the problem is with y_pred. You can concatenate it to X_test, transform it to a StandardDataset object, and then pass that one to the BinaryLabelDatasetMetric. The output object will have the methods for calculating different fairness metrics. I do not know how your dataset looks like, but here is a complete reproducible example that you can adapt to do this process for your dataset.