Merge pull request #27 from IBM/ColumnProjection

Extracting Numeric/String columns (Project operator)

demo/demo-lale-project.jl

@@ -0,0 +1,34 @@
+using Lale
+using DataFrames
+using CSV
+
+strfeat= laleoperator("Project","lale")(columns=Dict("type"=>"string"))
+numfeat = laleoperator("Project","lale")(columns=Dict("type"=>"number"))
+
+OneHotEncoder = laleoperator("OneHotEncoder")
+ConcatFeatures = laleoperator("ConcatFeatures","lale")
+Tree = laleoperator("DecisionTreeClassifier")
+KNN = laleoperator("KNeighborsClassifier")
+RF =  laleoperator("RandomForestClassifier")
+PCA = laleoperator("PCA")
+Hyperopt = laleoperator("Hyperopt","lale")
+
+df = CSV.read("./demo/old/credit.csv",DataFrame) 
+y = df[!,"class"] |> collect
+X = df[:,2:end]
+
+train_X,train_y,test_X,test_y = Lale.train_test_split(X, y,testprop=0.20)
+
+prep_to_numbers = 
+ ((stringprj >> OneHotEncoder(handle_unknown = "ignore")) & numericprj )>> ConcatFeatures
+planned_orig = prep_to_numbers >> ( Tree | KNN)
+lopt = LalePipeOptimizer(planned_orig,max_evals = 10,cv = 3)
+laletrained = fit(lopt,train_X,train_y)
+lalepred  = Lale.transform(laletrained,test_X)
+score(:accuracy,lalepred,test_y) |> println
+
+pca_tree_planned = prep_to_numbers >> PCA >> RF
+lopt = LalePipeOptimizer(pca_tree_planned,max_evals = 10,cv = 3)
+laletrained = fit(lopt,train_X,train_y)
+lalepred  = Lale.transform(laletrained,test_X)
+score(:accuracy,lalepred,test_y) |> println

demo/aif360.py → demo/old/aif360.py

@@ -3,7 +3,7 @@
 
 import pandas as pd
 pd.options.display.max_columns = None
-pd.concat([all_y, all_X], axis=1)
+df = pd.concat([all_y, all_X], axis=1)
 
 import lale.pretty_print
 lale.pretty_print.ipython_display(fairness_info)