model-training/preprocessing.py

import keras
print("Running Keras in version {}".format(keras.__version__))

import h5py
import numpy as np
import pandas as pd
import time
import sklearn.model_selection as sk
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
from imblearn.over_sampling import SMOTE
from imblearn.under_sampling import RandomUnderSampler
from imblearn.over_sampling import RandomOverSampler
from collections import Counter
import os
from sklearn.preprocessing import StandardScaler, MinMaxScaler

# preprocessing pipeline
# 

def Safelog(val):
    # get range of vector
    if val > 0:
        return np.log10(val)
    elif val < 0:
        return -np.log10(-val)
    else:
        return 0

def Safeexp(val):
    if val > 0:
        return -10 ** -val
    elif val < 0:
        return 10 ** val
    else:
        return 0
    

class FuncTransform():
    '''
    Class to transform and inverse transform data with given functions.
    Transform and inverse transform functions have to be given as dictionaries in the following format:
    {'key1': function1, 'key2': function2, ...}
    '''

    def __init__(self, func_transform, func_inverse):
        self.func_transform = func_transform
        self.func_inverse = func_inverse

    def fit(self, X, y=None):
        return self
    
    def transform(self, X, y=None):
        X = X.copy()
        for key in X.keys():   
            if "Class" not in key:
                X[key] = X[key].apply(self.func_transform[key])
        return X
    
    def fit_transform(self, X, y=None):
        self.fit(X)
        return self.transform(X, y)
    
    def inverse_transform(self, X_log):
        X_log = X_log.copy()
        for key in X_log.keys():
            if "Class" not in key:
                X_log[key] = X_log[key].apply(self.func_inverse[key])
        return X_log
    

def clustering(X, n_clusters=2, random_state=42, x_length=50, y_length=50):
    '''
    Function to cluster data with KMeans.
    '''

    class_labels = np.array([])
    grid_length = x_length * y_length
    iterations = int(len(X) / grid_length)

    for i in range(0, iterations):
        field = np.array(X['Barite'][(i*grid_length):(i*grid_length+grid_length)]
                         ).reshape(x_length, y_length)
        kmeans = KMeans(n_clusters=n_clusters, random_state=random_state).fit(
            field.reshape(-1, 1))

        class_labels = np.append(class_labels.astype(int), kmeans.labels_)
        
    if("Class" in X.columns and "Class" in X.columns):
        print("Class column already exists")
    else:
        class_labels_df = pd.DataFrame(class_labels, columns=['Class'])
        X_clustered = pd.concat([X, class_labels_df], axis=1)
    
    return X_clustered


def balancer(design, target, strategy, sample_fraction=0.5):
        
    number_features = (design.columns != "Class").sum()
    if("Class" not in design.columns):
        if("Class" in target.columns):
            classes = target['Class']
        else:
            raise Exception("No class column found")
    else:
        classes = design['Class']
        counter = classes.value_counts()
        print("Amount class 0 before:", counter[0] / (counter[0] + counter[1]) )
        print("Amount class 1 before:", counter[1] / (counter[0] + counter[1]) )
        df = pd.concat([design.loc[:,design.columns != "Class"], target.loc[:, target.columns != "Class"], classes], axis=1)
        
    if strategy == 'smote':
        print("Using SMOTE strategy")
        smote = SMOTE(sampling_strategy=sample_fraction)
        df_resampled, classes_resampled = smote.fit_resample(df.loc[:, df.columns != "Class"], df.loc[:, df.columns == "Class"])
        
    elif strategy == 'over':
        print("Using Oversampling")
        over = RandomOverSampler()
        df_resampled, classes_resampled = over.fit_resample(df.loc[:, df.columns != "Class"], df.loc[:, df.columns == "Class"])
        
    elif strategy == 'under':
        print("Using Undersampling")
        under = RandomUnderSampler()
        df_resampled, classes_resampled = under.fit_resample(df.loc[:, df.columns != "Class"], df.loc[:, df.columns == "Class"])

    else:
        classes_resampled = classes
        
    counter = classes_resampled["Class"].value_counts()
    print("Amount class 0 after:", counter[0] / (counter[0] + counter[1]) )
    print("Amount class 1 after:", counter[1] / (counter[0] + counter[1]) )
    
    design_resampled = pd.concat([df_resampled.iloc[:,0:number_features], classes_resampled], axis=1)
    target_resampled = pd.concat([df_resampled.iloc[:,number_features:], classes_resampled], axis=1)
    
    return design_resampled, target_resampled  


def plot_simulation(X, timestep, component='Barite', x_length=50, y_length=50):
    grid_length = x_length * y_length
    max_iter = int(len(X) / grid_length)
    if(timestep >= max_iter):
        raise Exception("timestep is not in the simulation range") 
    
    plt.imshow(np.array(X[component][(timestep*grid_length):(timestep*grid_length+grid_length)]).reshape(x_length,y_length), interpolation='bicubic', origin='lower')
    
    if("Class" in X.columns):
        plt.contour(np.array(X['Class'][(timestep*grid_length):(timestep*grid_length+grid_length)]).reshape(x_length,y_length), levels=[0.1], colors='red', origin='lower')
        
    plt.show()
    

def preprocessing(df_design, df_targets, func_dict_in, func_dict_out, sampling, test_size):
    
    df_design = clustering(df_design)
    
    df_design_log = FuncTransform(func_dict_in, func_dict_out).fit_transform(df_design)
    df_results_log = FuncTransform(func_dict_in, func_dict_out).fit_transform(df_targets)

    X_train, X_test, y_train, y_test = sk.train_test_split(df_design_log, df_results_log, test_size = test_size, random_state=42)

    X_train, y_train = balancer(X_train, y_train, sampling)
    
    scaler_X = MinMaxScaler()
    scaler_y = MinMaxScaler()
    
    X_train = scaler_X.fit_transform(X_train)
    X_test = scaler_X.transform(X_test)

    y_train = scaler_y.fit_transform(y_train)
    y_test = scaler_y.transform(y_test)
    
    X_train, X_val, y_train, y_val = sk.train_test_split(X_train, y_train, test_size = 0.1)
    
    return X_train, X_val, X_test, y_train, y_val, y_test