UholoDala's picture
Update app.py
8476caa
import gradio as gr
import pandas as pd
import joblib
import numpy as np
from sklearn.pipeline import Pipeline
from sklearn.impute import SimpleImputer
from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import StandardScaler, OneHotEncoder
from sklearn.linear_model import LogisticRegression
# Load the saved full pipeline from the file
full_pipeline = joblib.load('pipe.pkl')
# Define the predict function
def predict(gender, SeniorCitizen, Partner, Dependents, Contract, tenure, MonthlyCharges,
TotalCharges, PaymentMethod, PhoneService, MultipleLines, InternetService,
OnlineSecurity, OnlineBackup, DeviceProtection, TechSupport, StreamingTV,
StreamingMovies, PaperlessBilling):
# Create a DataFrame from the input data
input_data = pd.DataFrame({
'gender': [gender] if gender else ['Male'], # Replace None with default value
'SeniorCitizen': [SeniorCitizen] if SeniorCitizen is not None else [0], # Replace None with default value
'Partner': [Partner] if Partner else ['No'], # Replace None with default value
'Dependents': [Dependents] if Dependents else ['No'], # Replace None with default value
'tenure': [tenure] if tenure else [1], # Replace None with default value
'PhoneService': [PhoneService] if PhoneService else ['Yes'], # Replace None with default value
'MultipleLines': [MultipleLines] if MultipleLines else ['No'], # Replace None with default value
'InternetService': [InternetService] if InternetService else ['DSL'], # Replace None with default value
'OnlineSecurity': [OnlineSecurity] if OnlineSecurity else ['No'], # Replace None with default value
'OnlineBackup': [OnlineBackup] if OnlineBackup else ['No'], # Replace None with default value
'DeviceProtection': [DeviceProtection] if DeviceProtection else ['No'], # Replace None with default value
'TechSupport': [TechSupport] if TechSupport else ['No'], # Replace None with default value
'StreamingTV': [StreamingTV] if StreamingTV else ['No'], # Replace None with default value
'StreamingMovies': [StreamingMovies] if StreamingMovies else ['No'], # Replace None with default value
'Contract': [Contract] if Contract else ['Month-to-month'], # Replace None with default value
'PaperlessBilling': [PaperlessBilling] if PaperlessBilling else ['No'], # Replace None with default value
'PaymentMethod': [PaymentMethod] if PaymentMethod else ['Electronic check'], # Replace None with default value
'MonthlyCharges': [MonthlyCharges] if MonthlyCharges else [0.0], # Replace None with default value
'TotalCharges': [TotalCharges] if TotalCharges else [0.0] # Replace None with default value
})
# Make predictions using the loaded logistic regression model
#predict probabilities
predictions = full_pipeline.predict_proba(input_data)
#take the index of the maximum probability
index=np.argmax(predictions)
higher_pred_prob=round((predictions[0][index])*100)
#return predictions[0]
print(f'[Info] Predicted probabilities{predictions},{full_pipeline.classes_}')
if full_pipeline.classes_[index] == "Yes":
return f"This Customer is likely to Churn\nWe are {higher_pred_prob}% confident about this prediction"
else:
return f"This Customer is Not likely to Churn \nWe are {higher_pred_prob}% confident about this prediction"
# Setting Gradio App Interface
with gr.Blocks(css=".gradio-container {background-color: grey}",theme=gr.themes.Base(primary_hue='blue'),title='Uriel') as demo:
gr.Markdown("# Teleco Customer Churn Prediction #\n*This App allows the user to predict whether a customer will churn or not by entering values in the given fields. Any field left blank takes the default value.*")
# Receiving ALL Input Data here
gr.Markdown("**Demographic Data**")
with gr.Row():
gender = gr.Dropdown(label="Gender", choices=["Male", "Female"])
SeniorCitizen = gr.Radio(label="Senior Citizen", choices=[1, 0])
Partner = gr.Radio(label="Partner", choices=["Yes", "No"])
Dependents = gr.Radio(label="Dependents", choices=["Yes", "No"])
gr.Markdown("**Service Length and Charges (USD)**")
with gr.Row():
Contract = gr.Dropdown(label="Contract", choices=["Month-to-month", "One year", "Two year"])
tenure = gr.Slider(label="Tenure (months)", minimum=1, step=1, interactive=True)
MonthlyCharges = gr.Slider(label="Monthly Charges", step=0.05)
TotalCharges = gr.Slider(label="Total Charges", step=0.05)
# Phone Service Usage part
gr.Markdown("**Phone Service Usage**")
with gr.Row():
PhoneService = gr.Radio(label="Phone Service", choices=["Yes", "No"])
MultipleLines = gr.Dropdown(label="Multiple Lines", choices=[
"Yes", "No", "No phone service"])
# Internet Service Usage part
gr.Markdown("**Internet Service Usage**")
with gr.Row():
InternetService = gr.Dropdown(label="Internet Service", choices=["DSL", "Fiber optic", "No"])
OnlineSecurity = gr.Dropdown(label="Online Security", choices=["Yes", "No", "No internet service"])
OnlineBackup = gr.Dropdown(label="Online Backup", choices=["Yes", "No", "No internet service"])
DeviceProtection = gr.Dropdown(label="Device Protection", choices=["Yes", "No", "No internet service"])
TechSupport = gr.Dropdown(label="Tech Support", choices=["Yes", "No", "No internet service"])
StreamingTV = gr.Dropdown(label="TV Streaming", choices=["Yes", "No", "No internet service"])
StreamingMovies = gr.Dropdown(label="Movie Streaming", choices=["Yes", "No", "No internet service"])
# Billing and Payment part
gr.Markdown("**Billing and Payment**")
with gr.Row():
PaperlessBilling = gr.Radio(
label="Paperless Billing", choices=["Yes", "No"])
PaymentMethod = gr.Dropdown(label="Payment Method", choices=["Electronic check", "Mailed check", "Bank transfer (automatic)", "Credit card (automatic)"])
# Output Prediction
output = gr.Text(label="Outcome")
submit_button = gr.Button("Predict")
submit_button.click(fn= predict,
outputs= output,
inputs=[gender, SeniorCitizen, Partner, Dependents, Contract, tenure, MonthlyCharges, TotalCharges, PaymentMethod, PhoneService, MultipleLines, InternetService, OnlineSecurity, OnlineBackup, DeviceProtection, TechSupport, StreamingTV, StreamingMovies, PaperlessBilling],
),
# Add the reset and flag buttons
def clear():
output.value = ""
return 'Predicted values have been reset'
clear_btn = gr.Button("Reset", variant="primary")
clear_btn.click(fn=clear, inputs=None, outputs=output)
demo.launch(inbrowser = True)