P2M1_Deployment

app.py

@@ -0,0 +1,29 @@
+import streamlit as st
+import eda
+import prediction
+
+# Set Config dan icon
+st.set_page_config(
+        page_title='Churn Prediction',
+        layout='wide',
+        )
+
+# Hide Streamlit Style
+hide_streamlit_style = """
+            <style>
+            #MainMenu {visibility: hidden;}
+            footer {visibility: hidden;}
+            </style>
+            """
+st.markdown(hide_streamlit_style, unsafe_allow_html=True) 
+
+#Membuat navigasi
+st.sidebar.markdown("# Evan Derin Ihsanudin - RMT-FTDS-17")
+navigation = st.sidebar.selectbox('Pilih Halaman (Churn Prediction/EDA): ', ('Churn Prediction','Exploratory Data Analysis'))
+st.sidebar.image("https://imgur.com/t4aS0jH.png", use_column_width=True)
+
+#Run modul dengan if else
+if navigation == 'Churn Prediction' :
+    prediction.run()
+else :
+    eda.run()

churn_model.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7eea4b8ac89d86ecbbef7e8f38e1cd9c6d6b8822b4193f7d14791fe026e4b1e0
+size 262152

eda.py

@@ -0,0 +1,183 @@
+import streamlit as st
+import pandas as pd
+import seaborn as sns
+import matplotlib.pyplot as plt
+import plotly.express as px
+from PIL import Image
+
+
+
+def run() :
+    # Membuat Title 
+    st.markdown("<h1 style='text-align: center;'>Exploratory Data Analysis</h1>", unsafe_allow_html=True)
+    st.write('Berikut adalah EDA dari setiap feature')
+
+    # Import DF
+    df_eda = pd.read_csv('eda_churn.csv')
+
+    # Membuat Sub Header
+    st.subheader('**EDA Feature Churn**')
+    st.write('Dari visualisasi dibawah dapat disimpulkan bahwa :')
+    st.markdown('- *Customer* yang *churn* lebih banyak dari pada *customer* yang tidak *churn*')
+
+    # Membuat visualisasi Distribusi churn_risk_score
+    fig, ax =plt.subplots(1,2,figsize=(15,6))
+
+    sns.countplot(x='churn_risk_score', data=df_eda, palette="winter", ax=ax[0])
+    ax[0].set_xlabel("churn_risk_score", fontsize= 12)
+    ax[0].set_ylabel("# of Customer", fontsize= 12)
+    fig.suptitle('Customer Churn Distribution', fontsize=18, fontweight='bold')
+    ax[0].set_ylim(0,23000)
+    plt.xlabel("churn_risk_score", fontsize= 12)
+    plt.ylabel("# of Customer", fontsize= 12)
+    ax[0].set_xticks([0,1], ['Not Churn', 'Churn'])
+    for p in ax[0].patches:
+        ax[0].annotate("%.0f"%(p.get_height()), (p.get_x() + p.get_width() / 2,
+                        p.get_height()+405), ha='center', va='center',fontsize = 11) 
+
+    df_eda['churn_risk_score'].value_counts().plot(kind='pie', labels = ['Not Churn', 'Churn'],autopct='%1.1f%%', textprops = {"fontsize":12})
+    ax[1].set_ylabel("% of Customer", fontsize= 12)
+    st.pyplot(fig)
+
+    # Membuat Sub Header
+    st.subheader('**EDA Feature Age**')
+    st.write('Dari visualisasi dibawah dapat disimpulkan bahwa :')
+    st.markdown('- *Customer* paling banyak adalah *customer* yang memiliki *range* umur 40-50 tahun')
+    st.markdown('- *Customer* yang paling banyak *churn* adalah *customer* dengan *range* umur 50-60 tahun')
+    st.markdown('- Akan tetapi jika dilihat dari persentase *churn* pada setiap kelas *range* umur, maka tidak ada perbedaan signifikan')
+
+    #Visualisasi distribusi range age
+    fig, ax =plt.subplots(1,2,figsize=(15,6))
+    sns.countplot(x='AgeBin', data=df_eda, palette='winter', ax=ax[0])
+    ax[0].set_xlabel("Range Customer Age", fontsize= 12)
+    ax[0].set_ylabel("# of Customer", fontsize= 12)
+    fig.suptitle('Range Customer Age Distribution', fontsize=18, fontweight='bold')
+    ax[0].set_ylim(0,7600)
+    for p in ax[0].patches:
+        ax[0].annotate("%.0f"%(p.get_height()), (p.get_x() + p.get_width() / 2,
+                        p.get_height()+105), ha='center', va='center',fontsize = 10) 
+
+    df_eda['AgeBin'].value_counts().plot(kind='pie', autopct='%1.1f%%', textprops = {"fontsize":12})
+    ax[1].set_ylabel("% of Customer", fontsize= 10)
+    st.pyplot(fig)
+
+    # Membuat Visualisasi distribusi Age berdasarkan Churn
+    fig, ax =plt.subplots(1,2,figsize=(15,6))
+    sns.countplot(data = df_eda, x = 'AgeBin', hue="churn_risk_score", palette = 'winter', order = ['(10, 20]', '(20, 30]', '(30, 40]', '(40, 50]', '(50, 60]', '(60, 70]'], ax=ax[0])
+    ax[0].set_title('Range Age Distribution', fontsize=14, fontweight='bold',)
+    ax[0].set_xlabel("Range Age", fontsize= 12)
+    ax[0].set_ylabel("# of Customer", fontsize= 12)
+    ax[0].tick_params(axis="x", labelsize= 9.5)
+    ax[0].legend(fontsize=10,title='Churn Classification', loc='upper right', labels=['Not Churn', 'Churn'])
+    for p in ax[0].patches:
+        ax[0].annotate("%.0f"%(p.get_height()), (p.get_x() + p.get_width() / 2,
+                        p.get_height()+75), ha='center', va='center',fontsize = 10) 
+    ax[0].set_ylim(0,4700)
+
+    #Visualisasi % Churn dari setiap kelas
+    sns.barplot(x = 'AgeBin', y = 'churn_risk_score', data = df_eda, palette = 'winter', order = ['(10, 20]', '(20, 30]', '(30, 40]', '(40, 50]', '(50, 60]', '(60, 70]'],ax=ax[1])
+    ax[1].set_xlabel("Range Age", fontsize= 12)
+    ax[1].set_ylabel("% Churn", fontsize= 12)
+    ax[1].set_title('% Churn based on Age', fontsize=14, fontweight='bold')
+    ax[1].set_ylim(0,0.7)
+    for p in ax[1].patches:
+        ax[1].annotate("%.2f"%(p.get_height()), (p.get_x() + p.get_width() / 2,
+                        p.get_height()+0.03), ha='center', va='center',fontsize = 11) 
+    st.pyplot(fig)
+
+    # Membuat Sub Header
+    st.subheader('**EDA Feature Time Spent**')
+    st.write('Dari visualisasi dibawah dapat disimpulkan bahwa :')
+    st.markdown('- Jika dilihat pada visualisasi diatas maka `avg_time_spent` antara *customer* yang *churn* dan *customer* yang tidak *churn* tidak berbeda secara signifikan')
+
+    #  Visualisasi avg_time_spent vs Churn
+    fig =plt.figure(figsize=(15,6))
+    plt.rcParams['figure.figsize'] = (10, 5)
+    sns.boxenplot(y=df_eda['avg_time_spent'], x= df_eda['churn_risk_score'], palette = 'Blues')
+    plt.title('Average Time Spent vs Churn', fontsize = 20)
+    st.pyplot(fig)
+
+
+    # Membuat Sub Header
+    st.subheader('**EDA Feature Transaction Value**')
+    st.write('Dari visualisasi dibawah dapat disimpulkan bahwa :')
+    st.markdown('- *Customer* yang tidak *churn* memiliki *average transaction value* yang lebih tinggi (terpusat di 18.000-40.000) dari pada *customer* yang *churn* (terpusat di 16.000-36.000)')
+
+    #  Visualisasi avg_transaction_value vs Churn
+    fig =plt.figure(figsize=(15,6))
+    plt.rcParams['figure.figsize'] = (10, 5)
+    sns.boxenplot(y=df_eda['avg_transaction_value'], x= df_eda['churn_risk_score'], palette = 'Blues')
+    plt.title('Average Transaction Value vs Churn', fontsize = 20)
+    st.pyplot(fig)
+
+    # Membuat Sub Header
+    st.subheader('**EDA Feature Avg Frequency Login Days**')
+    st.write('Dari visualisasi dibawah dapat disimpulkan bahwa :')
+    st.markdown('- *Customer* yang tidak *churn* memiliki *average frequency login days* yang lebih rendah (terpusat di 8-20x) dari pada *customer* yang *churn* (terpusat di 10-25x)')
+
+    #  Visualisasi avg_frequency_login_days vs Churn
+    fig =plt.figure(figsize=(15,6))
+    plt.rcParams['figure.figsize'] = (10, 5)
+    sns.boxenplot(y=df_eda['avg_frequency_login_days'], x= df_eda['churn_risk_score'], palette = 'Blues')
+    plt.title('Average Frequency Login Days vs Churn', fontsize = 20)
+    st.pyplot(fig)
+
+    # Membuat Sub Header
+    st.subheader('**EDA Feature Point Wallet**')
+    st.write('Dari visualisasi dibawah dapat disimpulkan bahwa :')
+    st.markdown('- *Customer* yang tidak *churn* memiliki *points in wallet* yang lebih tinggi (terpusat di 700-800) dari pada *customer* yang *churn* (terpusat di 600-700)')
+
+    #  Visualisasi points_in_wallet vs Churn
+    fig =plt.figure(figsize=(15,6))
+    plt.rcParams['figure.figsize'] = (10, 5)
+    sns.boxenplot(y=df_eda['points_in_wallet'], x= df_eda['churn_risk_score'], palette = 'Blues')
+    plt.title('Points in Wallet vs Churn', fontsize = 20)
+    st.pyplot(fig)
+
+    # Membuat Sub Header
+    st.subheader('**EDA Feature Gender**')
+    st.write('Dari visualisasi dibawah dapat disimpulkan bahwa :')
+    st.markdown('- *Customer* paling banyak adalah *customer* wanita (50.1%). Akan tetapi tidak berbeda signifikan, hanya berbeda 0.1% dari *customer* pria')
+    st.markdown('- *Customer* yang paling banyak *churn* adalah *customer* wanita. Kemungkinan banyak pada kelas wanita karena *customer* paling banyak juga pada kelas ini')
+    st.markdown('- Akan tetapi jika dilihat dari persentase *churn* pada setiap kelas *gender*, maka tidak ada perbedaan signifikan')
+
+    #Visualisasi distribusi Gender
+    fig, ax =plt.subplots(1,2,figsize=(15,6))
+    sns.countplot(x='gender', data=df_eda, palette='winter', ax=ax[0])
+    ax[0].set_xlabel("Gender", fontsize= 12)
+    ax[0].set_ylabel("# of Customer", fontsize= 12)
+    fig.suptitle('Gender Distribution', fontsize=18, fontweight='bold')
+    ax[0].set_ylim(0,21000)
+    for p in ax[0].patches:
+        ax[0].annotate("%.0f"%(p.get_height()), (p.get_x() + p.get_width() / 2,
+                        p.get_height()+305), ha='center', va='center',fontsize = 10) 
+    df_eda['gender'].value_counts().plot(kind='pie', autopct='%1.1f%%', textprops = {"fontsize":12})
+    ax[1].set_ylabel("% of Customer", fontsize= 10)
+    st.pyplot(fig)
+
+    # Membuat Visualisasi distribusi Gender berdasarkan Churn
+    fig, ax =plt.subplots(1,2,figsize=(15,6))
+    sns.countplot(data = df_eda, x = 'gender', hue="churn_risk_score", palette = 'winter', ax=ax[0])
+    ax[0].set_title('Gender Distribution', fontsize=14, fontweight='bold',)
+    ax[0].set_xlabel("Gender", fontsize= 12)
+    ax[0].set_ylabel("# of Customer", fontsize= 12)
+    ax[0].tick_params(axis="x", labelsize= 9.5)
+    ax[0].legend(fontsize=10,title='Churn Classification', loc='upper right', labels=['Not Churn', 'Churn'])
+    for p in ax[0].patches:
+        ax[0].annotate("%.0f"%(p.get_height()), (p.get_x() + p.get_width() / 2,
+                        p.get_height()+175), ha='center', va='center',fontsize = 10) 
+    ax[0].set_ylim(0,13000)
+
+    #Visualisasi % Churn dari setiap kelas
+    sns.barplot(x = 'gender', y = 'churn_risk_score', data = df_eda, palette = 'winter',ax=ax[1])
+    ax[1].set_xlabel("Gender", fontsize= 12)
+    ax[1].set_ylabel("% Churn", fontsize= 12)
+    ax[1].set_title('% Churn based on Gender', fontsize=14, fontweight='bold')
+    ax[1].set_ylim(0,0.7)
+    for p in ax[1].patches:
+        ax[1].annotate("%.2f"%(p.get_height()), (p.get_x() + p.get_width() / 2,
+                        p.get_height()+0.02), ha='center', va='center',fontsize = 11) 
+    st.pyplot(fig)
+
+if __name__ == '__main__':
+    run()

prediction.py

@@ -0,0 +1,86 @@
+import streamlit as st
+import pandas as pd
+import numpy as np
+import pickle
+from tensorflow.keras.models import load_model
+
+
+def run() :
+  # Load Model
+  with open('preprocessor.pkl', 'rb') as file_2:
+    preprocessor = pickle.load(file_2)
+  model_churn = load_model('churn_model.h5', compile=False)
+
+  # Membuat Title 
+  st.markdown("<h1 style='text-align: center;'>Churn Prediction</h1>", unsafe_allow_html=True)
+
+  # Menambahkan Deskripsi Form
+  st.write('Page ini berisi model untuk memprediksi churn customer')
+
+  #Membuat Form
+  with st.form(key= 'form_customer'):
+      
+      st.markdown('### **Customer Data**')
+      user_id = st.text_input('User ID',value= '')
+      age = st.slider('Age',10,70,30)
+      gender = st.selectbox('Gender',('M','F'),index=1)
+      region_category = st.radio('Region', options=['City','Village','Town'], horizontal=True)
+      internet_option = st.selectbox('Internet Option',('Wi-Fi','Fiber_Optic', 'Mobile_Data'),index=1)
+      medium_of_operation = st.radio('Medium', options=['Desktop','Smartphone','Both'], horizontal=True)
+      st.markdown('---')
+      st.markdown('### **Login Data**')
+      days_since_last_login = st.slider('Days Since Last Login',0,30,3)
+      avg_frequency_login_days = st.slider('Avg Frequency Login Days',0,73,14)
+      st.markdown('---')
+      st.markdown('### **Membership Data**')
+      joined_through_referral = st.selectbox('Referral',('Yes','No'),index=1)
+      membership_category = st.selectbox('Membership Category',('No Membership','Basic Membership','Silver Membership', 'Premium Membership', 'Gold Membership', 'Platinum Membership'),index=1)
+      st.markdown('---')
+      st.markdown('### **Transaction Data**')
+      points_in_wallet = st.number_input('Points in Wallet', min_value=0, max_value=2070, value=600 ,step=1)
+      avg_transaction_value = st.number_input('Avg Transaction Value', min_value=800, max_value=90000, value=30000 ,step=1)
+      preferred_offer_types = st.radio('Offer Types', options=['Without Offers','Credit/Debit Card Offers','Gift Vouchers/Coupons'], horizontal=True)
+      used_special_discount = st.selectbox('Used Special Discount',('Yes','No'),index=1)
+      past_complaint = st.selectbox('Past Complaint',('Yes','No'),index=1)
+      feedback = st.selectbox('Feedback',('Poor Website','Poor Customer Service', 'Too many ads', 'Poor Product Quality', 'No reason specified', 'Products always in Stock', 'Reasonable Price', 'Quality Customer Care', 'User Friendly Website'),index=1)
+      submitted = st.form_submit_button('Predict')
+
+  # Create New Data 
+  data_inf = {
+      'user_id' : user_id,
+      'age' : age,
+      'gender' : gender,
+      'region_category' : region_category,
+      'internet_option' : internet_option, 
+      'medium_of_operation' : medium_of_operation,
+      'days_since_last_login' : days_since_last_login, 
+      'avg_frequency_login_days' : avg_frequency_login_days, 
+      'joined_through_referral' : joined_through_referral,
+      'membership_category' : membership_category,  
+      'points_in_wallet' : points_in_wallet, 
+      'avg_transaction_value' : avg_transaction_value,  
+      'used_special_discount' : used_special_discount,
+      'past_complaint' : past_complaint, 
+      'preferred_offer_types' : preferred_offer_types,                
+      'feedback' : feedback                                   
+  }
+
+  data_inf = pd.DataFrame([data_inf])
+  data_inf
+
+  if submitted :
+    # Feature Scaling and Feature Encoding
+    data_final = preprocessor.transform(data_inf)
+
+    # Predict using Linear Regression
+    y_inf_pred = np.where(model_churn.predict(data_final) >= 0.5, 1, 0)
+    
+    if y_inf_pred == 1:
+      prediction = 'Churn'
+    else:
+      prediction = 'Not Churn'
+
+    st.write('# Churn Prediction : ', prediction)
+
+if __name__ == '__main__':
+    run()

preprocessor.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ceb6d174d15913521943fd72ada7006de18e51eba93ef3d5a722538f5f8e4be
+size 2828