Update to synthetic data-set

app.py

@@ -76,7 +76,7 @@ with demo:
     with gr.Row():
         with gr.Column():
             gr.Markdown("### User")
-            bool_inputs = gr.CheckboxGroup(["Car", "Property", "Work phone", "Phone", "Email"], label="Which of the following do you actively hold or own?")
+            bool_inputs = gr.CheckboxGroup(["Car", "Property", "Mobile phone"], label="Which of the following do you actively hold or own?")
             num_children = gr.Slider(**CHILDREN_MIN_MAX, step=1, label="Number of children", info="How many children do you have ?")
             household_size = gr.Slider(**FAMILY_MIN_MAX, step=1, label="Household size", info="How many members does your household have? ?")
             total_income = gr.Slider(**INCOME_MIN_MAX, label="Income", info="What's you total yearly income (in euros) ?")
@@ -132,7 +132,7 @@ with demo:
         needing to decrypt any value.
 
         This server employs an [XGBoost](https://github.com/dmlc/xgboost) classifier model that has 
-        been trained on [this credit card data-set](https://www.kaggle.com/datasets/rikdifos/credit-card-approval-prediction/data).
+        been trained on a synthetic data-set.
         """
     )
 

backend.py

@@ -15,24 +15,29 @@ from settings import (
     CLIENT_FILES,
     SERVER_FILES,
     DEPLOYMENT_PATH,
-    INITIAL_INPUT_SHAPE,
+    PROCESSED_INPUT_SHAPE,
     INPUT_INDEXES,
     INPUT_SLICES,
     PRE_PROCESSOR_USER_PATH, 
+    PRE_PROCESSOR_BANK_PATH,
     PRE_PROCESSOR_THIRD_PARTY_PATH,
     CLIENT_TYPES,
     USER_COLUMNS,
+    BANK_COLUMNS,
     THIRD_PARTY_COLUMNS,
 )
 
 from utils.client_server_interface import MultiInputsFHEModelClient
 
 # Load pre-processor instances
-with PRE_PROCESSOR_USER_PATH.open('rb') as file:
-    PRE_PROCESSOR_USER = pickle.load(file)
-
-with PRE_PROCESSOR_THIRD_PARTY_PATH.open('rb') as file:
-    PRE_PROCESSOR_THIRD_PARTY = pickle.load(file)
+with (
+    PRE_PROCESSOR_USER_PATH.open('rb') as file_user, 
+    PRE_PROCESSOR_BANK_PATH.open('rb') as file_bank,
+    PRE_PROCESSOR_THIRD_PARTY_PATH.open('rb') as file_third_party,
+):
+    PRE_PROCESSOR_USER = pickle.load(file_user)
+    PRE_PROCESSOR_BANK = pickle.load(file_bank)
+    PRE_PROCESSOR_THIRD_PARTY = pickle.load(file_third_party)
 
 
 def shorten_bytes_object(bytes_object, limit=500):
@@ -124,8 +129,8 @@ def _send_to_server(client_id, client_type, file_name):
 
     Args:
         client_id (int): The client ID to consider.
-        client_type (Optional[str]): The type of client to consider (either 'user', 'bank', 'third_party' or 
-            None).
+        client_type (Optional[str]): The type of client to consider (either 'user', 'bank', 
+            'third_party' or None).
         file_name (str): File name to send (either 'evaluation_key' or 'encrypted_inputs').
     """
     # Get the paths to the encrypted inputs
@@ -213,7 +218,7 @@ def _encrypt_send(client_id, inputs, client_type):
     encrypted_inputs = client.quantize_encrypt_serialize_multi_inputs(
         inputs, 
         input_index=INPUT_INDEXES[client_type], 
-        initial_input_shape=INITIAL_INPUT_SHAPE, 
+        processed_input_shape=PROCESSED_INPUT_SHAPE, 
         input_slice=INPUT_SLICES[client_type],
     )
 
@@ -251,16 +256,12 @@ def pre_process_encrypt_send_user(client_id, *inputs):
     # Retrieve boolean values
     own_car = "Car" in bool_inputs
     own_property = "Property" in bool_inputs
-    work_phone = "Work phone" in bool_inputs
-    phone = "Phone" in bool_inputs
-    email = "Email" in bool_inputs
+    mobile_phone = "Mobile phone" in bool_inputs
 
     user_inputs = pandas.DataFrame({
         "Own_car": [own_car],
         "Own_property": [own_property],
-        "Work_phone": [work_phone],
-        "Phone": [phone],
-        "Email": [email],
+        "Mobile_phone": [mobile_phone],
         "Num_children": [num_children],
         "Household_size": [household_size],
         "Total_income": [total_income],
@@ -291,8 +292,16 @@ def pre_process_encrypt_send_bank(client_id, *inputs):
             the encrypted input to send. 
     """
     account_age = inputs[0]
+
+    bank_inputs = pandas.DataFrame({
+        "Account_age": [account_age],
+    })
+
+    bank_inputs = bank_inputs.reindex(BANK_COLUMNS, axis=1)
+
+    preprocessed_bank_inputs = PRE_PROCESSOR_BANK.transform(bank_inputs)
     
-    return _encrypt_send(client_id, account_age, "bank")
+    return _encrypt_send(client_id, preprocessed_bank_inputs, "bank")
     
 
 def pre_process_encrypt_send_third_party(client_id, *inputs):
@@ -421,5 +430,4 @@ def decrypt_output(client_id):
     # Determine the predicted class
     output = numpy.argmax(output_proba, axis=1).squeeze()
     
-    # A "0" output means approving the credit card has low risk, while "1" is high risk
-    return "Credit card is likely to be approved ✅" if output == 0 else "Credit card is likely to be denied ❌"
+    return "Credit card is likely to be approved ✅" if output == 1 else "Credit card is likely to be denied ❌"

data/gpt_data.csv

@@ -0,0 +1,57 @@
+ID,Car,Property,Work phone,Phone,Email,Number of children,Household size,Income,Age,Income type,Education,Family,Occupation,Housing,Account age (months),Employed,Years of employment,Credit card Approval
+0,Yes,Yes,No,Yes,Yes,2,4,50000,35,Salaried,Higher education,Married,Engineer,House / apartment,36,Yes,5,Yes
+1,No,No,Yes,Yes,No,0,1,20000,50,Pensioner,Secondary,Widow,Retired,Rented apartment,12,No,0,No
+2,Yes,Yes,No,No,Yes,1,3,75000,40,Self-employed,Higher education,Civil marriage,Business Owner,House / apartment,48,Yes,10,Yes
+3,No,Yes,Yes,No,No,1,2,30000,28,Salaried,Secondary,Single,Teacher,With parents,24,Yes,3,Yes
+4,Yes,No,No,Yes,Yes,0,1,45000,32,Self-employed,Higher education,Divorced,Freelancer,Rented apartment,60,Yes,7,Yes
+5,No,Yes,Yes,Yes,No,3,5,35000,45,Salaried,Incomplete higher,Married,Clerk,House / apartment,30,Yes,10,Yes
+6,Yes,No,No,No,Yes,0,1,25000,55,Pensioner,Lower secondary,Widow,Retired,House / apartment,120,No,0,No
+7,No,Yes,Yes,Yes,Yes,2,3,60000,37,Salaried,Higher education,Civil marriage,Manager,House / apartment,36,Yes,12,Yes
+8,Yes,Yes,Yes,Yes,Yes,0,1,80000,29,Self-employed,Higher education,Single,Entrepreneur,House / apartment,48,Yes,6,Yes
+9,No,Yes,No,Yes,No,4,6,40000,38,Salaried,Secondary,Married,Salesperson,With parents,60,Yes,15,Yes
+10,Yes,No,Yes,No,Yes,1,2,55000,26,Self-employed,Higher education,Civil marriage,Designer,Rented apartment,12,Yes,3,Yes
+11,No,No,No,No,No,0,1,30000,60,Pensioner,Lower secondary,Widow,Retired,House / apartment,180,No,0,No
+12,Yes,Yes,Yes,Yes,Yes,2,4,70000,45,Salaried,Higher education,Married,Manager,House / apartment,72,Yes,20,Yes
+13,No,Yes,Yes,Yes,Yes,3,5,65000,50,Salaried,Incomplete higher,Divorced,Teacher,House / apartment,96,Yes,25,Yes
+14,Yes,No,Yes,Yes,No,0,1,85000,33,Self-employed,Higher education,Single,Consultant,Rented apartment,36,Yes,10,Yes
+15,No,Yes,No,No,Yes,1,3,32000,55,Pensioner,Secondary,Widow,Retired,House / apartment,150,No,0,No
+16,Yes,Yes,Yes,Yes,Yes,2,4,90000,41,Salaried,Higher education,Married,Doctor,House / apartment,60,Yes,15,Yes
+17,No,No,No,Yes,Yes,0,1,23000,24,Student,Incomplete higher,Single,Student,With parents,6,No,0,No
+18,Yes,No,Yes,No,Yes,1,3,48000,35,Self-employed,Secondary,Divorced,Chef,Rented apartment,24,Yes,5,Yes
+19,No,Yes,No,Yes,No,3,5,55000,50,Salaried,Higher education,Widow,Engineer,House / apartment,120,Yes,20,Yes
+20,Yes,No,Yes,Yes,Yes,0,2,70000,30,Salaried,Higher education,Civil marriage,Lawyer,Rented apartment,36,Yes,8,Yes
+21,No,Yes,Yes,No,Yes,2,4,32000,40,Self-employed,Secondary,Married,Carpenter,House / apartment,48,Yes,12,Yes
+22,Yes,No,No,Yes,No,1,2,26000,29,Student,Higher education,Single,Student,With parents,12,No,0,No
+23,No,Yes,Yes,Yes,Yes,0,1,45000,55,Pensioner,Lower secondary,Divorced,Retired,House / apartment,180,No,0,No
+24,Yes,Yes,No,Yes,Yes,1,3,80000,46,Salaried,Higher education,Married,Architect,House / apartment,72,Yes,18,Yes
+25,No,No,Yes,No,No,2,5,37000,33,Self-employed,Secondary,Single,Plumber,With parents,24,Yes,6,Yes
+26,Yes,Yes,Yes,Yes,No,0,1,65000,28,Self-employed,Higher education,Single,Graphic Designer,Rented apartment,36,Yes,4,Yes
+27,No,No,No,No,Yes,1,2,29000,53,Pensioner,Incomplete higher,Widow,Retired,House / apartment,144,No,0,No
+28,Yes,Yes,Yes,No,Yes,3,6,58000,39,Salaried,Secondary,Divorced,Teacher,House / apartment,60,Yes,10,Yes
+29,No,Yes,No,Yes,No,0,1,40000,31,Salaried,Higher education,Single,Nurse,Rented apartment,48,Yes,7,Yes
+30,Yes,No,Yes,Yes,Yes,2,4,51000,42,Self-employed,Higher education,Civil marriage,Business Owner,House / apartment,60,Yes,14,Yes
+31,No,Yes,No,No,No,1,3,34000,47,Salaried,Secondary,Married,Factory Worker,With parents,72,Yes,9,No
+32,No,No,Yes,Yes,Yes,0,1,85000,25,Self-employed,Higher education,Single,Entrepreneur,Rented apartment,12,Yes,2,Yes
+33,Yes,Yes,No,No,No,3,5,30000,55,Salaried,Secondary,Married,Teacher,House / apartment,240,Yes,30,Yes
+34,No,No,Yes,Yes,Yes,1,2,45000,40,Salaried,Higher education,Divorced,Nurse,With parents,36,Yes,10,Yes
+35,Yes,No,No,Yes,No,0,1,20000,50,Pensioner,Lower secondary,Widow,Retired,Rented apartment,60,No,0,No
+36,No,Yes,Yes,No,Yes,2,4,35000,30,Salaried,Higher education,Civil marriage,Clerk,House / apartment,24,Yes,5,Yes
+37,Yes,No,Yes,Yes,No,1,3,60000,33,Self-employed,Secondary,Single,Freelancer,Rented apartment,48,Yes,7,No
+38,No,No,No,No,Yes,0,1,75000,28,Salaried,Higher education,Single,Manager,With parents,12,Yes,3,Yes
+39,Yes,Yes,Yes,Yes,No,3,6,40000,45,Salaried,Incomplete higher,Married,Salesperson,House / apartment,96,Yes,15,Yes
+40,Yes,No,Yes,Yes,Yes,0,1,30000,65,Pensioner,Higher education,Widow,Retired,House / apartment,420,No,40,Yes
+41,No,Yes,No,No,No,1,2,22000,26,Salaried,Secondary,Single,Junior Clerk,With parents,24,Yes,2,No
+42,Yes,No,Yes,Yes,No,2,4,18000,35,Salaried,Lower secondary,Married,Factory Worker,House / apartment,60,Yes,10,Yes
+43,No,No,No,Yes,Yes,0,1,50000,55,Self-employed,Higher education,Divorced,Consultant,Rented apartment,180,Yes,30,Yes
+44,Yes,Yes,Yes,No,No,3,5,25000,40,Salaried,Incomplete higher,Married,Cook,House / apartment,120,Yes,15,No
+45,No,Yes,No,Yes,Yes,1,2,35000,28,Self-employed,Secondary,Single,Graphic Designer,Rented apartment,36,Yes,5,Yes
+46,Yes,No,Yes,No,Yes,0,1,45000,30,Salaried,Higher education,Civil marriage,Engineer,With parents,48,Yes,8,Yes
+47,No,Yes,Yes,Yes,No,2,3,27000,48,Salaried,Secondary,Divorced,Salesperson,House / apartment,240,Yes,20,No
+48,Yes,No,Yes,Yes,No,2,4,40000,45,Self-employed,Secondary,Married,Artist,House / apartment,120,Yes,15,Yes
+49,No,Yes,No,No,Yes,0,1,55000,50,Salaried,Higher education,Widow,Scientist,Rented apartment,300,No,25,Yes
+50,Yes,No,Yes,Yes,Yes,1,2,26000,27,Salaried,Incomplete higher,Single,Junior Accountant,With parents,36,Yes,3,No
+51,No,No,No,No,No,3,6,22000,38,Salaried,Lower secondary,Divorced,Bus Driver,House / apartment,72,Yes,10,Yes
+52,Yes,Yes,Yes,No,Yes,0,1,32000,55,Pensioner,Higher education,Married,Retired,House / apartment,420,No,30,No
+53,No,Yes,No,Yes,No,2,3,29000,33,Self-employed,Secondary,Civil marriage,Shop Owner,Rented apartment,60,Yes,7,Yes
+54,Yes,No,Yes,Yes,Yes,1,2,68000,29,Salaried,Higher education,Single,Software Developer,With parents,48,Yes,6,Yes
+55,No,No,No,No,Yes,0,1,36000,42,Self-employed,Incomplete higher,Divorced,Freelancer,House / apartment,96,Yes,12,No

deployment_files/client.zip

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3a407200f51839a85032c3f69ca3413e9cdb99d2e03722f03450599d2f8d318d
-size 76138
+oid sha256:859c5895ff365bfd7fa8b592784717d14094c29aaee032c2f23716d225c15855
+size 29354

deployment_files/pre_processor_bank.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c76179d6cb576a61f6d57e29f45545d8c4dd2d86a3818730f5c6bdd35faebe4a
+size 1098

deployment_files/pre_processor_third_party.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:223d27a2af60a06845da0191cf725b7bab7f31ebb0a84bc530cbd32001ac4f47
-size 1588
+oid sha256:8bc9d12a6ac40e3b7f88f3f15e688879b819bc4a995ad0e91fc08af85f55513b
+size 1128

deployment_files/pre_processor_user.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:60f9d3d6e98f3eadc7a30340dae414df3f42fabd73da01fe0a6ece91186964b5
-size 6222
+oid sha256:a4300414ee222355b2ac9e0f5e4dc21b50f0a4de8e17e0ed4121dc766231d010
+size 3340

deployment_files/server.zip

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:98a6a55f0be2a5fc9b2174b9a9686550fe2a89c3c2ae937aeb31e4057f645a03
-size 3419
+oid sha256:3490a405f8634b3c95f6a2d3fb8cf0276e3f49adc4b25911afab7f97524c2f7a
+size 2308

development.py

@@ -5,16 +5,12 @@ import numpy
 import pandas
 import pickle
 
-from sklearn.model_selection import train_test_split
-from sklearn.metrics import accuracy_score
-from imblearn.over_sampling import SMOTE
-
 from settings import (
-    DEPLOYMENT_PATH, 
-    RANDOM_STATE, 
+    DEPLOYMENT_PATH,
     DATA_PATH, 
     INPUT_SLICES, 
     PRE_PROCESSOR_USER_PATH, 
+    PRE_PROCESSOR_BANK_PATH,
     PRE_PROCESSOR_THIRD_PARTY_PATH,
     USER_COLUMNS,
     BANK_COLUMNS,
@@ -34,19 +30,12 @@ def get_processed_multi_inputs(data):
 
 print("Load and pre-process the data")
 
-# Original data set can be found here : 
-# https://www.kaggle.com/datasets/rikdifos/credit-card-approval-prediction/data
-# It then has been cleaned using the following notebook : 
-# https://www.kaggle.com/code/samuelcortinhas/credit-cards-data-cleaning
-# A few additional pre-processing steps has bee applied to this data set as well : 
-# - "ID" column has been removed 
-# - "Total_income" values have been multiplied by 0.14 to make its median match France's annual
-#    salary one from 2023 (22050 euros) 
+# Load the data
 data = pandas.read_csv(DATA_PATH, encoding="utf-8")
 
 # Define input and target data
 data_x = data.copy()
-data_y = data_x.pop("Target").copy()
+data_y = data_x.pop("Target").copy().to_frame()
 
 # Get data from all parties
 data_user = data_x[USER_COLUMNS].copy()
@@ -54,30 +43,22 @@ data_bank = data_x[BANK_COLUMNS].copy()
 data_third_party = data_x[THIRD_PARTY_COLUMNS].copy()
 
 # Feature engineer the data
-pre_processor_user, pre_processor_third_party = get_pre_processors()
+pre_processor_user, pre_processor_bank, pre_processor_third_party = get_pre_processors()
 
 preprocessed_data_user = pre_processor_user.fit_transform(data_user)
-preprocessed_data_bank = data_bank.to_numpy()
+preprocessed_data_bank = pre_processor_bank.fit_transform(data_bank)
 preprocessed_data_third_party = pre_processor_third_party.fit_transform(data_third_party)
 
 preprocessed_data_x = numpy.concatenate((preprocessed_data_user, preprocessed_data_bank, preprocessed_data_third_party), axis=1)
 
-# The initial data-set is very imbalanced: use SMOTE to get better results
-x, y = SMOTE().fit_resample(preprocessed_data_x, data_y)
-
-# Retrieve the training and testing data
-X_train, X_test, y_train, y_test = train_test_split(
-    x, y, stratify=y, test_size=0.3, random_state=RANDOM_STATE
-)
-
 
 print("\nTrain and compile the model")
 
-model = MultiInputXGBClassifier(max_depth=3, n_estimators=40)
+model = MultiInputXGBClassifier(max_depth=3, n_estimators=20)
 
-model, sklearn_model = model.fit_benchmark(X_train, y_train)
+model, sklearn_model = model.fit_benchmark(preprocessed_data_x, data_y)
  
-multi_inputs_train = get_processed_multi_inputs(X_train)
+multi_inputs_train = get_processed_multi_inputs(preprocessed_data_x)
 
 model.compile(*multi_inputs_train, inputs_encryption_status=["encrypted", "encrypted", "encrypted"])
 
@@ -86,19 +67,6 @@ if DEPLOYMENT_PATH.is_dir():
     shutil.rmtree(DEPLOYMENT_PATH)
 
 
-print("\nEvaluate the models")
-
-y_pred_sklearn = sklearn_model.predict(X_test)
-
-print(f"Sklearn accuracy score : {accuracy_score(y_test, y_pred_sklearn )*100:.2f}%")
-
-multi_inputs_test = get_processed_multi_inputs(X_test)
-
-y_pred_simulated = model.predict_multi_inputs(*multi_inputs_test, simulate=True)
-
-print(f"Concrete ML accuracy score (simulated) : {accuracy_score(y_test, y_pred_simulated)*100:.2f}%")
-
-
 print("\nSave deployment files")
 
 # Save files needed for deployment (and enable cross-platform deployment)
@@ -106,10 +74,13 @@ fhe_dev = MultiInputsFHEModelDev(DEPLOYMENT_PATH, model)
 fhe_dev.save(via_mlir=True)
 
 # Save pre-processors
-with PRE_PROCESSOR_USER_PATH.open('wb') as file:
-    pickle.dump(pre_processor_user, file)
-
-with PRE_PROCESSOR_THIRD_PARTY_PATH.open('wb') as file:
-    pickle.dump(pre_processor_third_party, file)
+with (
+    PRE_PROCESSOR_USER_PATH.open('wb') as file_user, 
+    PRE_PROCESSOR_BANK_PATH.open('wb') as file_bank,
+    PRE_PROCESSOR_THIRD_PARTY_PATH.open('wb') as file_third_party,
+):
+    pickle.dump(pre_processor_user, file_user)
+    pickle.dump(pre_processor_bank, file_bank)
+    pickle.dump(pre_processor_third_party, file_third_party)
 
 print("\nDone !")

requirements.txt

@@ -1,3 +1,2 @@
 gradio==3.40.1
-concrete-ml==1.3.0
-imbalanced-learn==0.11.0
+concrete-ml==1.3.0

settings.py

@@ -14,6 +14,7 @@ SERVER_FILES = REPO_DIR / "server_files"
 
 # Path targeting pre-processor saved files
 PRE_PROCESSOR_USER_PATH = DEPLOYMENT_PATH / 'pre_processor_user.pkl'
+PRE_PROCESSOR_BANK_PATH = DEPLOYMENT_PATH / 'pre_processor_bank.pkl'
 PRE_PROCESSOR_THIRD_PARTY_PATH = DEPLOYMENT_PATH / 'pre_processor_third_party.pkl'
 
 # Create the necessary directories
@@ -25,13 +26,10 @@ SERVER_FILES.mkdir(exist_ok=True)
 SERVER_URL = "http://localhost:8000/" 
 
 # Path to data file
-# Details about pre-processing steps can be found in the 'development.py' and 'pre_processing.py'
-# files
 DATA_PATH = "data/data.csv"
 
 # Development settings
-RANDOM_STATE = 0
-INITIAL_INPUT_SHAPE = (1, 49)
+PROCESSED_INPUT_SHAPE = (1, 39)
 
 CLIENT_TYPES = ["user", "bank", "third_party"]
 INPUT_INDEXES = {
@@ -40,13 +38,13 @@ INPUT_INDEXES = {
     "third_party": 2,
 }
 INPUT_SLICES = {
-    "user": slice(0, 42),  # First position: start from 0
-    "bank": slice(42, 43),  # Second position: start from n_feature_user
-    "third_party": slice(43, 49),  # Third position: start from n_feature_user + n_feature_bank
+    "user": slice(0, 36),  # First position: start from 0
+    "bank": slice(36, 37),  # Second position: start from n_feature_user
+    "third_party": slice(37, 39),  # Third position: start from n_feature_user + n_feature_bank
 }
 
 USER_COLUMNS = [
-    'Own_car', 'Own_property', 'Work_phone', 'Phone', 'Email', 'Num_children', 'Household_size', 
+    'Own_car', 'Own_property', 'Mobile_phone', 'Num_children', 'Household_size', 
     'Total_income', 'Age', 'Income_type', 'Education_type', 'Family_status', 'Housing_type', 
     'Occupation_type',
 ]

utils/client_server_interface.py

@@ -27,9 +27,15 @@ class MultiInputsFHEModelClient(FHEModelClient):
 
         super().__init__(*args, **kwargs)
     
-    def quantize_encrypt_serialize_multi_inputs(self, x: numpy.ndarray, input_index, initial_input_shape, input_slice) -> bytes:
+    def quantize_encrypt_serialize_multi_inputs(
+        self, 
+        x: numpy.ndarray, 
+        input_index, 
+        processed_input_shape, 
+        input_slice
+    ) -> bytes:
 
-        x_padded = numpy.zeros(initial_input_shape)
+        x_padded = numpy.zeros(processed_input_shape)
 
         x_padded[:, input_slice] = x
 

utils/pre_processing.py

@@ -1,16 +1,9 @@
-"""Data pre-processing functions.
-
-The pre-processing steps are heavily inspired by the following notebook :
-https://www.kaggle.com/code/rikdifos/credit-card-approval-prediction-using-ml
-
-Additional steps, mostly including renaming some values or features, were added for better user 
-experience.
-"""
+"""Data pre-processing functions."""
 
 import numpy
 from sklearn.compose import ColumnTransformer
 from sklearn.pipeline import Pipeline
-from sklearn.preprocessing import OneHotEncoder, FunctionTransformer, KBinsDiscretizer
+from sklearn.preprocessing import OneHotEncoder, FunctionTransformer, StandardScaler
 
 
 def _get_pipeline_replace_one_hot(func, value):
@@ -23,10 +16,6 @@ def _get_pipeline_replace_one_hot(func, value):
         ("one_hot", OneHotEncoder(),),
     ])
 
-
-def _replace_values_geq(column, value):
-    return numpy.where(column >= value, f"{value}_or_more", column)
-
 def _replace_values_eq(column, value):
     for desired_value, values_to_replace in value.items():
         column = numpy.where(numpy.isin(column, values_to_replace), desired_value, column)
@@ -35,41 +24,32 @@ def _replace_values_eq(column, value):
 def get_pre_processors():
     pre_processor_user = ColumnTransformer(
         transformers=[
-            (
-                "replace_num_children",
-                _get_pipeline_replace_one_hot(_replace_values_geq, 2),
-                ['Num_children']
-            ),
-            (
-                "replace_household_size",
-                _get_pipeline_replace_one_hot(_replace_values_geq, 3),
-                ['Household_size']
-            ),
-            (
-                "replace_income_type",
-                _get_pipeline_replace_one_hot(_replace_values_eq, {"Public Sector": ["Retired", "Student"]}),
-                ['Income_type']
-            ),
-            (
-                "replace_education_type",
-                _get_pipeline_replace_one_hot(_replace_values_eq, {"Higher education": ["Academic degree"]}),
-                ['Education_type']
-            ),
             (
                 "replace_occupation_type_labor",
                 _get_pipeline_replace_one_hot(
                     _replace_values_eq, 
                         {
-                            "Labor_work": ["Cleaning staff", "Cooking staff", "Drivers", "Laborers", "Low-wage laborers", "Security staff", "Waiters/barmen staff"],
-                            "Office_work": ["Accountants", "Core staff", "HR staff", "Medicine staff", "Private service staff", "Realty agents", "Sales staff", "Secretaries"],
-                            "High_tech_work": ["Managers", "High skill tech staff", "IT staff"],
+                            "Labor_work": [
+                                "Cooking Staff", "Carpenter", "Plumber", "Factory Worker", "Bus Driver"
+                            ],
+                            "Office_work": [
+                                "Business Owners", "Office Worker", "Accountant", "Entrepreneur", "Salesperson"
+                            ],
+                            "High_tech_work": ["Engineer", "Manager", "Consultant", "Software Developer"],
                         },
                 ),
                 ['Occupation_type']
             ),
-            ('one_hot_housing_fam_status', OneHotEncoder(), ['Housing_type', 'Family_status']),
-            ('qbin_total_income', KBinsDiscretizer(n_bins=3, strategy='quantile', encode="onehot"), ['Total_income']),
-            ('bin_age', KBinsDiscretizer(n_bins=5, strategy='uniform', encode="onehot"), ['Age']),
+            ('one_hot_others', OneHotEncoder(), ['Housing_type', 'Family_status', 'Education_type', 'Income_type']),
+            ('standard_scaler', StandardScaler(), ['Num_children', 'Household_size', 'Total_income', 'Age']),
+        ],
+        remainder='passthrough',
+        verbose_feature_names_out=False,
+    )
+
+    pre_processor_bank = ColumnTransformer(
+        transformers=[
+            ('standard_scaler', StandardScaler(), ['Account_age']),
         ],
         remainder='passthrough',
         verbose_feature_names_out=False,
@@ -77,10 +57,10 @@ def get_pre_processors():
 
     pre_processor_third_party = ColumnTransformer(
         transformers=[
-            ('bin_years_employed', KBinsDiscretizer(n_bins=5, strategy='uniform', encode="onehot"), ['Years_employed'])
+            ('standard_scaler', StandardScaler(), ['Years_employed']),
         ],
         remainder='passthrough',
         verbose_feature_names_out=False,
     )
 
-    return pre_processor_user, pre_processor_third_party
+    return pre_processor_user, pre_processor_bank, pre_processor_third_party