Commit Deployment Final

README.md

@@ -0,0 +1,12 @@
+---
+title: Laptop Price Prediction
+emoji: 🌍
+colorFrom: indigo
+colorTo: purple
+sdk: streamlit
+sdk_version: 1.21.0
+app_file: app.py
+pinned: false
+---
+
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,29 @@
+import streamlit as st
+import eda
+import prediction
+from streamlit_option_menu import option_menu
+
+
+st.sidebar.header("HaiMeds Customer Churn Prediction")
+
+
+with st.sidebar:
+    st.write("Ediashta Revindra - FTDS-020")
+    selected = option_menu(
+        "Menu",
+        [
+            "Distribution",
+            "Correlation Matrix",
+            "Regression",
+        ],
+        icons=["bar-chart", "link-45deg", "code-square"],
+        menu_icon="cast",
+        default_index=0,
+    )
+
+if selected == "Distribution":
+    eda.distribution()
+elif selected == "Correlation Matrix":
+    eda.corr_matrix()
+elif selected == "Regression":
+    prediction.predict()

column_transformer.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25e90257db3cb67cf0c961fac606fbc5a90ee7a79b02d6a4329abe4c53667783
+size 3983

eda.py

@@ -0,0 +1,194 @@
+import streamlit as st
+import pandas as pd
+import seaborn as sns
+import matplotlib.pyplot as plt
+
+st.set_page_config(
+    page_title="Laptop Price Regression",
+    layout="wide",
+    initial_sidebar_state="expanded",
+)
+
+# dataset
+dataset = "https://raw.githubusercontent.com/ediashta/p2-ftds020-rmt-m1/main/churn.csv"
+data = pd.read_csv(dataset)
+
+
+def distribution():
+    # distribution plot
+    st.title("HaiMeds Customer Distribution")
+    col1, col2 = st.columns(2)
+
+    hist_plot_1 = col1.selectbox(
+        "Choose Table",
+        ("Age", "Last Login (Days)", "Avg. Time Spent"),
+    )
+    hist_plot(hist_plot_1, col1)
+
+    hist_plot_2 = col2.selectbox(
+        "Choose Table",
+        ("Avg. Transaction", "Avg. Login Frequency (Days)", "Points"),
+    )
+    hist_plot(hist_plot_2, col2)
+
+    col1, col2 = st.columns(2)
+    bar_plot_1 = col1.selectbox(
+        "Choose Table",
+        ("Gender", "Region", "Membership", "Referral", "Preferred Offer", "Devices"),
+    )
+    bar_plot(bar_plot_1, col1)
+
+    bar_plot_2 = col2.selectbox(
+        "Choose Table",
+        (
+            "Internet",
+            "Used Discount",
+            "Offer Application Preference",
+            "Past Complaint",
+            "Complaint Status",
+            "Feedback",
+        ),
+    )
+    bar_plot(bar_plot_2, col2)
+
+    st.subheader("Churn Risk Score Distribution")
+    churn_score()
+
+
+def corr_matrix():
+    # distribution plot
+    st.title("Features Correlation")
+    col1, col2 = st.columns([7, 5])
+
+    # correlation for numerical
+    fig = plt.figure(figsize=(10, 10))
+    corr_matrix = data[
+        [
+            "age",
+            "days_since_last_login",
+            "avg_time_spent",
+            "avg_transaction_value",
+            "avg_frequency_login_days",
+            "points_in_wallet",
+            "churn_risk_score",
+        ]
+    ].corr(method="spearman")
+    sns.heatmap(corr_matrix, annot=True, cmap="mako", square=True)
+    plt.xticks(rotation=45)
+    plt.yticks(rotation=45)
+    col1.pyplot(fig)
+
+    feature_importance_info = """
+        **Feature Importance:**
+
+        - **gender:** 0.0
+        - **region_category:** 0.0223
+        - **membership_category:** 0.7859
+        - **joining_date:** 0.0
+        - **joined_through_referral:** 0.0355
+        - **preferred_offer_types:** 0.0434
+        - **medium_of_operation:** 0.0218
+        - **internet_option:** 0.0025
+        - **last_visit_time:** 0.0604
+        - **used_special_discount:** 0.0092
+        - **offer_application_preference:** 0.0179
+        - **past_complaint:** 0.0072
+        - **complaint_status:** 0.0054
+        - **feedback:** 0.4561
+        """
+    col2.markdown(feature_importance_info)
+
+
+def bar_plot(var, col):
+    # ram storage dist
+    col.write("Distribusi " + var + " terbanyak")
+    var_old = var
+
+    if var == "Gender":
+        var = "gender"
+    elif var == "Region":
+        var = "region_category"
+    elif var == "Membership":
+        var = "membership_category"
+    elif var == "Referral":
+        var = "joined_through_referral"
+    elif var == "Preferred Offer":
+        var = "preferred_offer_types"
+    elif var == "Devices":
+        var = "medium_of_operation"
+    elif var == "Internet":
+        var = "internet_option"
+    elif var == "Used Discount":
+        var = "used_special_discount"
+    elif var == "Offer Application Preference":
+        var = "offer_application_preference"
+    elif var == "Past Complaint":
+        var = "past_complaint"
+    elif var == "Complaint Status":
+        var = "complaint_status"
+    elif var == "Feedback":
+        var = "feedback"
+
+    fig = plt.figure(figsize=(10, 5))
+    ax1 = sns.countplot(
+        data=data,
+        x=var,
+        palette="mako",
+    )
+    plt.xlabel(var_old)
+    ax1.bar_label(container=ax1.containers[0], labels=data[var].value_counts().values)
+    col.pyplot(fig)
+
+
+def hist_plot(var, col):
+    # check price distribution
+    col.write("Distribusi " + var)
+    var_old = var
+
+    if var == "Age":
+        var = "age"
+    elif var == "Last Login (Days)":
+        var = "days_since_last_login"
+    elif var == "Avg. Time Spent":
+        var = "avg_time_spent"
+    elif var == "Avg. Transaction":
+        var = "avg_transaction_value"
+    elif var == "Avg. Login Frequency (Days)":
+        var = "avg_frequency_login_days"
+    elif var == "Points":
+        var = "points_in_wallet"
+    else:
+        var = var
+
+    fig = plt.figure(figsize=(10, 5))
+
+    palette = sns.color_palette("mako_r", 50)
+    plt.xlabel(var_old)
+    plot = sns.histplot(data=data, x=var, kde=True, bins=50, color="teal")
+
+    for bin_, i in zip(plot.patches, palette):
+        bin_.set_facecolor(i)
+
+    col.pyplot(fig)
+
+
+def churn_score():
+    fig = plt.figure(figsize=(20, 5))
+    plt.ylabel("Churn Risk Score")
+
+    sorted_scores = data["churn_risk_score"].value_counts().sort_index(ascending=False)
+    ax = sns.countplot(
+        data=data, y="churn_risk_score", palette="mako", order=sorted_scores.index
+    )
+    # Get the value counts for each category of 'churn_risk_score'
+    value_counts = data["churn_risk_score"].value_counts()
+
+    # Add labels on top of each bar
+    for idx, count in enumerate(value_counts):
+        ax.text(count + 5, idx, str(count), va="center")
+
+    st.pyplot(fig)
+
+
+if __name__ == "__main__":
+    distribution()

prediction.py

@@ -0,0 +1,183 @@
+import streamlit as st
+import pandas as pd
+import numpy as np
+import pickle
+from tensorflow.keras.models import load_model
+
+
+# load file
+
+with open("./column_transformer.pkl", "rb") as file_1:
+    column_transformer = pickle.load(file_1)
+
+model_functional = load_model("./functional_model.keras")
+
+
+def predict():
+    # form
+    with st.form("key=churn_prediction"):
+        st.subheader("Churn Score Prediction")
+
+        st.markdown("**Customer Data**")
+
+        col1, col2 = st.columns(2, gap="large")
+        age = col1.number_input(label="Age", help="Customer Age", step=1, value=20)
+
+        membership = col2.selectbox(
+            label="Membership Category",
+            options=(
+                "No Membership",
+                "Basic Membership",
+                "Premium Membership",
+                "Silver Membership",
+                "Gold Membership",
+                "Platinum Membership",
+            ),
+        )
+        st.markdown("---")
+
+        col1, col2, col3, col4 = st.columns(4, gap="large")
+        region = col1.radio(
+            label="Region",
+            help="Customer Residence Region",
+            options=("Town", "City", "Village"),
+        )
+
+        referral = col2.radio(
+            label="Referral", help="Joined Through Referral?", options=("Yes", "No")
+        )
+
+        device = col3.radio(
+            label="Device(s)",
+            help="Device Used",
+            options=("Smartphone", "Desktop", "Both"),
+        )
+
+        internet = col4.radio(
+            label="Internet Connection", options=("Wi-Fi", "Fiber_Optic", "Mobile_Data")
+        )
+
+        st.markdown("---")
+
+        st.markdown("**Customer Behavior**")
+        col1, col2, col3, col4, col5 = st.columns(5, gap="large")
+
+        last_login = col1.number_input(
+            label="Last Login", help="Days Since Last Login", step=1, value=6
+        )
+
+        avg_time = col2.number_input(
+            label="Avg. Usage Time", help="Average Usage Time (Minutes)", value=30
+        )
+
+        avg_login = col3.number_input(
+            label="Avg. Login Frequency",
+            help="Average Login Frequency (Days)",
+            value=14,
+        )
+
+        points = col4.number_input(label="Points in Wallet", value=300)
+
+        transaction = col5.number_input(label="Avg. Transaction", value=100, help="USD")
+
+        st.markdown("---")
+
+        col1, col2, col3 = st.columns(3, gap="large")
+
+        offer_pref = col1.selectbox(
+            label="Preferred Offer Type",
+            options=(
+                "Gift Vouchers/Coupons",
+                "Credit/Debit Card Offers",
+                "Without Offers",
+            ),
+        )
+
+        used_disc = col2.radio(label="Used Discount Before?", options=("Yes", "No"))
+
+        offer_app = col3.radio(
+            label="Application Preference Offer?", options=("Yes", "No")
+        )
+
+        st.markdown("---")
+        col1, col2, col3 = st.columns(3, gap="large")
+
+        complaints = col1.radio(label="Past Complaint?", options=("Yes", "No"))
+
+        complaints_status = col2.selectbox(
+            label="Complaint Status",
+            options=(
+                "Not Appllicable",
+                "Unsolved",
+                "Solved",
+                "Solved in Follow-up",
+                "No Information Available",
+            ),
+        )
+        feedback = col3.selectbox(
+            label="Feedback Type", options=("Neutral", "Positive", "Negative")
+        )
+        submitted = st.form_submit_button("Predict")
+
+    # inferencing
+    data_inf = [
+        {
+            "age": age,
+            "region_category": region,
+            "membership_category": membership,
+            "joined_through_referral": referral,
+            "preferred_offer_types": offer_pref,
+            "medium_of_operation": device,
+            "internet_option": internet,
+            "days_since_last_login": last_login,
+            "avg_time_spent": avg_time,
+            "avg_transaction_value": transaction,
+            "avg_frequency_login_days": avg_login,
+            "points_in_wallet": points,
+            "used_special_discount": used_disc,
+            "offer_application_preference": offer_app,
+            "past_complaint": complaints,
+            "complaint_status": complaints_status,
+            "feedback": feedback,
+        }
+    ]
+
+    data_inf = pd.DataFrame(data_inf)
+
+    st.dataframe(data_inf)
+
+    data_inf_transform = column_transformer.transform(data_inf)
+    y_pred_inf = model_functional.predict(data_inf_transform)
+    y_pred_inf = np.where(y_pred_inf >= 0.65, 1, 0)
+
+    st.write("Prediksi Churn Pelanggan Tersebut adalah :")
+    if y_pred_inf[0] == 1:
+        html_str = f"""
+                    <style>
+                    p.a {{
+                    font: bold 36px Arial;
+                    color: teal;
+                    }}
+                    </style>
+                    <p class="a">Pelanggan Tidak Berpotensi Churn</p>
+                    """
+        st.markdown(html_str, unsafe_allow_html=True)
+        st.write(
+            "Dapat menekankan program loyalty agar pelanggan tetap menggunakan layanan"
+        )
+    else:
+        html_str = f"""
+                    <style>
+                    p.a {{
+                    font: bold 36px Arial;
+                    color: red;
+                    }}
+                    </style>
+                    <p class="a">Pelanggan Berpotensi Churn</p>
+                    """
+        st.markdown(html_str, unsafe_allow_html=True)
+        st.write("Dapat diberikan promosi untuk menarik pelanggan kembali")
+
+
+if __name__ == "__main__":
+    predict()

requirements.txt

@@ -0,0 +1,10 @@
+seaborn
+pandas
+matplotlib
+plotly
+Pillow
+scikit-learn==1.2.2
+category_encoders
+xgboost
+streamlit_option_menu
+tensorflow