import streamlit as st import pandas as pd import numpy as np import joblib import matplotlib.pyplot as plt from sklearn.decomposition import PCA st.set_page_config(page_title="KMeans User Segmentation", layout="centered") st.title("👥 KMeans User Segmentation") st.write("Upload a CSV or enter values manually to assign users into **KMeans clusters**.") # ----------------------------- # Load artifacts # ----------------------------- @st.cache_resource def load_artifacts(): model = joblib.load("src/kmeans_user_segmentation.pkl") scaler = joblib.load("src/scaler.pkl") return model, scaler model, scaler = load_artifacts() st.success("Model + scaler loaded successfully ✅") # ----------------------------- # Feature columns used in training # ----------------------------- FEATURES = [ "Average Screen Time", "Average Spent on App (INR)", "Ratings", "Last Visited Minutes", "New Password Request", ] # ----------------------------- # Sidebar options # ----------------------------- st.sidebar.header("⚙️ Options") input_mode = st.sidebar.radio("Input method", ["Upload CSV", "Manual Input"]) show_pca = st.sidebar.checkbox("Show PCA 2D plot (based on input data)", value=True) # ----------------------------- # Helper: validate + prepare # ----------------------------- def prepare_features(df: pd.DataFrame) -> pd.DataFrame: missing = [c for c in FEATURES if c not in df.columns] if missing: raise ValueError(f"Missing columns: {missing}") X = df[FEATURES].copy() # Make sure numeric for c in FEATURES: X[c] = pd.to_numeric(X[c], errors="coerce") if X.isna().any().any(): bad_cols = X.columns[X.isna().any()].tolist() raise ValueError(f"Non-numeric or missing values found in: {bad_cols}. Please clean your CSV.") return X # ----------------------------- # Input: CSV # ----------------------------- df = None if input_mode == "Upload CSV": uploaded = st.file_uploader("Upload CSV", type=["csv"]) if uploaded is None: st.info("Upload a CSV to start.") else: df = pd.read_csv(uploaded) st.subheader("📄 Data Preview") st.dataframe(df.head()) # ----------------------------- # Input: Manual # ----------------------------- else: st.info("Manual input is best for quick testing. For real segmentation, use CSV upload.") n = st.slider("Number of users (rows)", 1, 20, 5) rows = [] for i in range(n): st.markdown(f"### User {i+1}") c1, c2 = st.columns(2) screen_time = c1.number_input("Average Screen Time", min_value=0.0, value=10.0, key=f"st_{i}") spent = c2.number_input("Average Spent on App (INR)", min_value=0.0, value=100.0, key=f"sp_{i}") c3, c4 = st.columns(2) ratings = c3.number_input("Ratings", min_value=0.0, value=7.0, key=f"ra_{i}") last_visit = c4.number_input("Last Visited Minutes", min_value=0.0, value=1000.0, key=f"lv_{i}") pwd_req = st.number_input("New Password Request", min_value=0.0, value=1.0, key=f"pw_{i}") rows.append([screen_time, spent, ratings, last_visit, pwd_req]) df = pd.DataFrame(rows, columns=FEATURES) st.subheader("📄 Manual Data") st.dataframe(df) # ----------------------------- # Predict # ----------------------------- if df is not None and len(df) > 0: try: X = prepare_features(df) X_scaled = scaler.transform(X) clusters = model.predict(X_scaled) out = df.copy() out["cluster"] = clusters st.subheader("✅ Cluster Results") st.dataframe(out.head(20)) st.subheader("📊 Cluster Distribution") st.write(out["cluster"].value_counts().sort_index()) # Cluster profile (means) st.subheader("📌 Cluster Profile (mean values)") profile = out.groupby("cluster")[FEATURES].mean(numeric_only=True).round(2) st.dataframe(profile) # Simple bar chart for profile st.subheader("📈 Average Feature Values per Cluster") fig, ax = plt.subplots(figsize=(9, 4)) profile.plot(kind="bar", ax=ax) ax.set_xlabel("Cluster") ax.set_ylabel("Mean value") ax.legend(loc="best") st.pyplot(fig) # PCA plot (optional) if show_pca and len(out) >= 3: pca = PCA(n_components=2, random_state=42) X_pca = pca.fit_transform(X_scaled) st.subheader("🧭 PCA 2D Projection (Input Data)") fig2, ax2 = plt.subplots(figsize=(6, 5)) sc = ax2.scatter(X_pca[:, 0], X_pca[:, 1], c=clusters) ax2.set_title("KMeans Clusters (PCA Projection)") ax2.set_xlabel("PC1") ax2.set_ylabel("PC2") st.pyplot(fig2) # Download results st.subheader("⬇️ Download") csv = out.to_csv(index=False).encode("utf-8") st.download_button("Download results CSV", csv, file_name="clustered_users.csv", mime="text/csv") st.info("Tip: Clustering is unsupervised — clusters are groups with similar behavior patterns, not 'correct/incorrect' labels.") except Exception as e: st.error(str(e)) st.stop()