capstone_gradio_app_embedding.py

# -*- coding: utf-8 -*-
"""Capstone Gradio App Embedding.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1zsT_lHGVHzG29XSb4tMF3UdA6glyWnRx
"""

from google.colab import drive
drive.mount('/content/drive')

!pip install gradio

"""### **DATA PREP**"""

import pandas as pd
# # Specify the correct file paths for your data files
# test = pd.read_csv('/content/drive/MyDrive/datasets capstone/test.csv')
# y_train = pd.read_csv("/content/drive/MyDrive/datasets capstone/y_train.csv")
# X_train =pd.read_csv('/content/drive/MyDrive/datasets capstone/X_train.csv')

# #X_test = pd.read_csv("/content/test.csv")

path = "/content/drive/MyDrive/Capstone Project /"

train = pd.read_csv(path + 'Train.csv')

# use lambda function to remove \t make our model more robst
train = train.applymap(lambda x: x.replace("\t" , '' ) if isinstance (x , str) else x)
# "  " , " "
train = train.applymap(lambda x: x.replace("  " , ' ' ) if isinstance (x , str) else x)

# drop what we don't need
train.drop(columns=['MRG', 'user_id', 'ZONE1', 'ZONE2', 'TOP_PACK'], inplace=True)

train.head(1)

train["REGION"].fillna(method='ffill', inplace=True)
train["TENURE"].fillna(method='ffill', inplace=True)
train["MONTANT"].fillna(train["MONTANT"].median(), inplace=True)
train["FREQUENCE_RECH"].fillna(0, inplace=True)
train["REVENUE"].fillna(train["REVENUE"].median(), inplace=True)
train["ARPU_SEGMENT"].fillna(0, inplace=True)
train["FREQUENCE"].fillna(0, inplace=True)
train["DATA_VOLUME"].fillna(0, inplace=True)
train["ON_NET"].fillna(0, inplace=True)
train["ORANGE"].fillna(0, inplace=True)
train["TIGO"].fillna(0, inplace=True)
# train["ZONE1"].fillna(0, inplace=True)
# train["ZONE2"].fillna(0, inplace=True)
# train["MRG"].fillna(method='ffill', inplace=True)
train["REGULARITY"].fillna(train["REGULARITY"].mean(), inplace=True)
#train["TOP_PACK"].fillna(method='ffill', inplace=True)
train["FREQ_TOP_PACK"].fillna(train["FREQ_TOP_PACK"].mean(), inplace=True)

train['TENURE'] = train['TENURE'].str.replace('D 3-6 month', '1', regex=True)
train['TENURE'] = train['TENURE'].str.replace('E 6-9 month', '2', regex=True)
train['TENURE'] = train['TENURE'].str.replace('F 9-12 month', '3', regex=True)
train['TENURE'] = train['TENURE'].str.replace('J 21-24 month', '4', regex=True)
train['TENURE'] = train['TENURE'].str.replace('G 12-15 month', '5', regex=True)
train['TENURE'] = train['TENURE'].str.replace('H 15-18 month', '6', regex=True)
train['TENURE'] = train['TENURE'].str.replace('I 18-21 month', '7', regex=True)
train['TENURE'] = train['TENURE'].str.replace('K > 24 month', '8', regex=True)

# train['TENURE'].value_counts()

# Define a dictionary to map values
region_mapping = {
    'DAKAR': '1',
    'THIES': '2',
    'SAINT-LOUIS': '3',
    'LOUGA': '4',
    'KAOLACK': '5',
    'DIOURBEL': '6',
    'TAMBACOUNDA': '7',
    'KAFFRINE': '8',
    'KOLDA': '9',
    'FATICK': '10',
    'ZIGUINCHOR': '11',
    'SEDHIOU': '12',
    'KEDOUGOU': '13',
    'MATAM' : '14'
}

# Use the replace method to map values
train['REGION'] = train['REGION'].replace(region_mapping)

# Look at the new value_counts
# print(train['REGION'].value_counts())

"""## **FITTING AND TRAINING**"""

# train.head(1)

"""Select target and features"""

y = train['CHURN']

x = train.drop(columns='CHURN', axis=1)

# y.head(3)

# x.head(2)

from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(x,y,test_size = 0.5,random_state=45 )# , stratify=y)

#Further split X_train and y_train into train and validation sets
X_train,X_val,y_train,y_val = train_test_split(X_train,y_train,test_size = 0.3, random_state=1 )#, stratify=y)

"""### SCALE NUMERICAL COLUMNS"""

num_cols = ['MONTANT', 'FREQUENCE_RECH', 'REVENUE', 'ARPU_SEGMENT', 'FREQUENCE',
       'DATA_VOLUME', 'ON_NET', 'ORANGE', 'TIGO',
       'REGULARITY', 'FREQ_TOP_PACK']

scaler = StandardScaler()

X_train[num_cols] = scaler.fit_transform(X_train[num_cols])

X_val[num_cols] = scaler.fit_transform(X_val[num_cols])

#X_train.head(3)

"""### ENCODE CATEGORICAL COLS WITH NUMERICAL VALUES WITH MANY N_UNIQUE( ) VALS

"""

!pip install category_encoders

# import category_encoders as ce

# encoder_ = ce.SumEncoder(cols=['TOP_PACK'])

# encoder.fit(x, y)
# X_train = encoder.transform(X_train)

# X_val = encoder.transform(X_val)

# X_test = encoder.transform(X_test)

from sklearn.model_selection import train_test_split, cross_val_score
from sklearn.metrics import accuracy_score, confusion_matrix, recall_score, precision_recall_curve, f1_score
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import ExtraTreesRegressor
from sklearn.preprocessing import LabelEncoder

# Create an instance
model = ExtraTreesRegressor(
    n_estimators=100,  # Number of trees in the forest
    max_depth=10,  # Maximum depth of the tree
    random_state=42  # Random seed for reproducibility
)

# Train the model
MODEL = model.fit(X_train, y_train)

"""## **Check if our model is working**"""

y_pred = MODEL.predict(X_test)

y_pred

"""Since our model is working correctly , Inspect the X_test features , to be used as the user input to interact with the model"""

X_test.head()



def classifier_1(result):
    if result > 0.9:
        return "Customer will churn"
    else:
        return "Customer will not churn"

def predict(REGION,TENURE	, MONTANT ,	FREQUENCE_RECH, REVENUE	, ARPU_SEGMENT ,FREQUENCE	, DATA_VOLUME	, ON_NET, ORANGE ,	TIGO, REGULARITY ,FREQ_TOP_PACK):
    input_array = np.array([[REGION,TENURE	, MONTANT ,	FREQUENCE_RECH, REVENUE	, ARPU_SEGMENT ,FREQUENCE	, DATA_VOLUME	, ON_NET, ORANGE ,	TIGO, REGULARITY ,FREQ_TOP_PACK]])

    pred = MODEL.predict(input_array)

    output = classifier_1 (pred[0])

    if output == "Customer will churn":
      return [(0, output)]
    else :
      return [(1, output)]

"""Check if the function will work"""

predict(1,2,3,4,5,6,7,8,9,1,1,1,1)

variable_definitions

X_test['FREQ_TOP_PACK'].min() , X_test['FREQ_TOP_PACK'].max()

import gradio as gr

#tenure = tenure_dropdown
REGION = gr.inputs.Slider(minimum=1, maximum=13, label='Location of each client')
TENURE = gr.inputs.Slider(minimum=1, maximum=8, label="Duration in network")
MONTANT = gr.inputs.Slider(minimum=22, maximum=470000, label="Top up amount")
FREQUENCE_RECH = gr.inputs.Slider(minimum=1, maximum=131, label="income frequency")
REVENUE = gr.inputs.Slider(minimum=1, maximum=532177, label="ARPU_SEGMENT")
ARPU_SEGMENT = gr.inputs.Slider(minimum=1, maximum= 177392, label="FREQUENCE")
FREQUENCE = gr.inputs.Slider(minimum=1, maximum=91, label="DATA_VOLUME")
DATA_VOLUME =gr.inputs.Slider(minimum=0, maximum=1702309, label="ON_NET")
ON_NET = gr.inputs.Slider(minimum=0, maximum=36687, label="ORANGE")
ORANGE = gr.inputs.Slider(minimum=0, maximum= 6721, label="TIGO")
TIGO = gr.inputs.Slider(minimum=0, maximum=4174, label="ZONE1")
REGULARITY = gr.inputs.Slider(minimum=1, maximum=62, label="ZONE2")
FREQ_TOP_PACK = gr.inputs.Slider(minimum=1, maximum= 592, label="REGULARITY")


op = gr.outputs.HighlightedText(color_map={"Customer will churn":"pink", "Customer will not churn":"yellow"})

gr.Interface(predict , inputs = [REGION,TENURE, MONTANT ,	FREQUENCE_RECH, REVENUE	, ARPU_SEGMENT ,FREQUENCE	, DATA_VOLUME	, ON_NET, ORANGE ,TIGO, REGULARITY ,FREQ_TOP_PACK], outputs=op,
             live = True).launch(debug=True)

import gradio as gr

# Input sliders
REGION = gr.inputs.Slider(minimum=1, maximum=13, label='Location of each client')
TENURE = gr.inputs.Slider(minimum=1, maximum=8, label="Duration in network")
MONTANT = gr.inputs.Slider(minimum=22, maximum=470000, label="Top-up amount")
FREQUENCE_RECH = gr.inputs.Slider(minimum=1, maximum=131, label="Income frequency")
REVENUE = gr.inputs.Slider(minimum=1, maximum=532177, label="ARPU_SEGMENT")
ARPU_SEGMENT = gr.inputs.Slider(minimum=1, maximum=177392, label="FREQUENCE")
FREQUENCE = gr.inputs.Slider(minimum=1, maximum=91, label="DATA_VOLUME")
DATA_VOLUME = gr.inputs.Slider(minimum=0, maximum=1702309, label="ON_NET")
ON_NET = gr.inputs.Slider(minimum=0, maximum=36687, label="ORANGE")
ORANGE = gr.inputs.Slider(minimum=0, maximum=6721, label="TIGO")
TIGO = gr.inputs.Slider(minimum=0, maximum=4174, label="ZONE1")
REGULARITY = gr.inputs.Slider(minimum=1, maximum=62, label="ZONE2")
FREQ_TOP_PACK = gr.inputs.Slider(minimum=1, maximum=592, label="REGULARITY")

# Output configuration
op = gr.outputs.HighlightedText(color_map={"Customer will churn": "pink", "Customer will not churn": "yellow"})

# Create and launch the interface
gr.Interface(predict, inputs=[REGION, TENURE, MONTANT, FREQUENCE_RECH, REVENUE, ARPU_SEGMENT, FREQUENCE,
                              DATA_VOLUME, ON_NET, ORANGE, TIGO, REGULARITY, FREQ_TOP_PACK], outputs=op,
             live=False).launch(debug=False)

# # Map numerical values to labels
# tenure_labels = {
#     0: "3-6 months",
#     1: "6-9 months",
#     2: "9-12 months",
#     3: "12-15 months",
#     4: "15-18 months",
#     5: "18-21 months",
#     6: "21-24 months",
#     7: "> 24 months"
# }

# # Reverse the mapping for predictions
# tenure_values = {v: k for k, v in tenure_labels.items()}

# # Create a dropdown menu with labels
# tenure_dropdown = gr.inputs.Dropdown(list(tenure_labels.values()), label="TENURE")