src/streamlit_app.py

import streamlit as st
import pandas as pd
import numpy as np

st.set_page_config(page_title="SPK Pemilihan Cabang untuk Lokasi Coffeeshop", layout="wide")

st.title("📊 Sistem Pendukung Keputusan Pemilihan Cabang untuk Lokasi Coffeeshop")
st.markdown("Metode: AHP, TOPSIS, dan Profile Matching")

# --- Step 1: Pilih Metode ---
st.sidebar.header("⚙ Pengaturan")
method = st.selectbox("🧠 Pilih Metode", ["None","TOPSIS", "Profile Matching", "AHP"], key="selected_method")

criteria_default = "Building Area, Road Access, Distance, Rental Price"
criteria_input = st.sidebar.text_area("Masukkan Kriteria (pisahkan dengan koma)", criteria_default)
criteria = [c.strip() for c in criteria_input.split(",") if c.strip()]

alternatives_default = "Location 1, Location 2, Location 3, Location 4"
alternatives_input = st.sidebar.text_area("Masukkan Alternatif (pisahkan dengan koma)", alternatives_default)
alternatives = [a.strip() for a in alternatives_input.split(",") if a.strip()]

if not criteria or not alternatives:
    st.warning("Masukkan minimal satu kriteria dan alternatif.")
    st.stop()

# ---------------------------
if method == "AHP":
    st.subheader("🔗 Perbandingan Berpasangan Antar Kriteria (AHP)")

    pairwise_criteria_data = np.zeros((len(criteria), len(criteria)))
    for i in range(len(criteria)):
        for j in range(i + 1, len(criteria)):
            pairwise_criteria_data[i, j] = 1.0

    pairwise_criteria_df = pd.DataFrame(pairwise_criteria_data, index=criteria, columns=criteria)

    edited_pairwise_criteria_matrix = st.data_editor(
        pairwise_criteria_df,
        use_container_width=True,
        key="pairwise_comparison_criteria_matrix",
        hide_index=False,
    )

    # Reciprocity for criteria matrix
    final_pairwise_criteria_matrix = edited_pairwise_criteria_matrix.copy()
    for i in range(len(criteria)):
        for j in range(i + 1, len(criteria)):
            val = final_pairwise_criteria_matrix.iloc[i, j]
            if val == 0:
                val = 1e-9
            final_pairwise_criteria_matrix.iloc[j, i] = 1 / val
        final_pairwise_criteria_matrix.iloc[i, i] = 1.0

    # Hitung bobot kriteria AHP
    weights_ahp_criteria = pd.Series(dtype='float64')
    if not final_pairwise_criteria_matrix.apply(pd.to_numeric, errors='coerce').isnull().values.any():
        try:
            numeric_criteria_matrix = final_pairwise_criteria_matrix.apply(pd.to_numeric)
            norm_matrix = numeric_criteria_matrix / numeric_criteria_matrix.sum()
            weights_ahp_criteria = norm_matrix.mean(axis=1)
            weights_ahp_criteria /= weights_ahp_criteria.sum()
            
            st.write("🎯 Bobot Kriteria dari AHP:")
            st.dataframe(weights_ahp_criteria.round(3).to_frame(name="Bobot AHP"))

            RI = {1: 0.00, 2: 0.00, 3: 0.58, 4: 0.90, 5: 1.12, 6: 1.24, 7: 1.32, 8: 1.41, 9: 1.45, 10: 1.49}
            n_criteria = len(criteria)
            # Weighted Sum Vector (WSV)
            wsv = numeric_criteria_matrix.dot(weights_ahp_criteria)
            # λ_max manual: rata-rata dari WSV[i] / W[i]
            lambda_max_criteria = (wsv / weights_ahp_criteria).mean()
            # CI manual
            CI_criteria = (lambda_max_criteria - n_criteria) / (n_criteria - 1) if n_criteria > 1 else 0

            CR_criteria = np.nan
            if n_criteria in RI and RI[n_criteria] > 0:
                CR_criteria = CI_criteria / RI[n_criteria]

            st.subheader("✅ Analisis Konsistensi Kriteria AHP")
            st.write(f"$\\lambda_{{max}} = {lambda_max_criteria:.10f}$")
            st.write(f"CI = {CI_criteria:.10f}")
            if not np.isnan(CR_criteria):
                st.write(f"CR = {CR_criteria:.10f}")
                if CR_criteria < 0.1:
                    st.success("Konsistensi Rasio Kriteria AHP *BAIK* (< 0.1)")
                else:
                    st.error("Konsistensi Rasio Kriteria AHP *BURUK* (>= 0.1). Harap sesuaikan perbandingan kriteria.")
            else:
                st.info("CR kriteria tidak dapat dihitung (untuk 1 kriteria atau RI tidak tersedia).")

        except Exception as e:
            st.error(f"Terjadi kesalahan saat menghitung bobot kriteria AHP: {e}")
            st.warning("Pastikan semua nilai pada matriks perbandingan kriteria adalah numerik.")
    else:
        st.warning("Matriks perbandingan kriteria belum lengkap atau tidak valid.")

    # --- Perbandingan berpasangan antar alternatif untuk setiap kriteria ---
    st.subheader("🔗 Perbandingan Berpasangan Antar Alternatif untuk Setiap Kriteria (AHP)")

    alternative_pairwise_matrices = {}
    alternative_weights_per_criterion = {}

    for criterion in criteria:
        st.markdown(f"### Kriteria: *{criterion}*")

        pairwise_alt_data = np.zeros((len(alternatives), len(alternatives)))
        for i in range(len(alternatives)):
            for j in range(i + 1, len(alternatives)):
                pairwise_alt_data[i, j] = 1.0

        pairwise_alt_df = pd.DataFrame(pairwise_alt_data, index=alternatives, columns=alternatives)

        st.markdown(f"Isi perbandingan berpasangan alternatif untuk *{criterion}* (misal: 3 berarti alternatif baris 3x lebih baik dari alternatif kolom terhadap {criterion}).")
        st.markdown("Nilai akan otomatis disesuaikan secara resiprokal saat perhitungan.")

        edited_pairwise_alt_matrix = st.data_editor(
            pairwise_alt_df,
            use_container_width=True,
            key=f"pairwise_comparison_alternatives_matrix_{criterion}",
            hide_index=False,
        )

        final_pairwise_alt_matrix = edited_pairwise_alt_matrix.copy()
        for i in range(len(alternatives)):
            for j in range(i + 1, len(alternatives)):
                val = final_pairwise_alt_matrix.iloc[i, j]
                if val == 0:
                    val = 1e-9
                final_pairwise_alt_matrix.iloc[j, i] = 1 / val
            final_pairwise_alt_matrix.iloc[i, i] = 1.0

        alternative_pairwise_matrices[criterion] = final_pairwise_alt_matrix

        # Hitung bobot alternatif untuk criterion ini
        try:
            numeric_alt_matrix = final_pairwise_alt_matrix.apply(pd.to_numeric)
            norm_matrix = numeric_alt_matrix / numeric_alt_matrix.sum()
            weights_alternatives = norm_matrix.mean(axis=1)
            weights_alternatives /= weights_alternatives.sum()
            alternative_weights_per_criterion[criterion] = weights_alternatives
            st.write(f"Bobot alternatif untuk kriteria *{criterion}*:")
            st.dataframe(weights_alternatives.round(3).to_frame(name=f"Bobot alternatif ({criterion})"))
        except Exception as e:
            st.error(f"Error menghitung bobot alternatif pada kriteria {criterion}: {e}")

elif method == "None":
    st.subheader("Pilih Metode terlebih dahulu")
elif method =="TOPSIS":
    # TOPSIS 
    st.subheader("📋 Input Nilai Alternatif terhadap Kriteria")
    empty_data = pd.DataFrame(np.nan, index=alternatives, columns=criteria)
    df = st.data_editor(empty_data, use_container_width=True, key="input_matrix")
    if df.isnull().values.any():
        st.warning("⚠ Harap lengkapi semua nilai pada tabel sebelum menjalankan perhitungan untuk TOPSIS/Profile Matching.")

    criteria_default_type= {
        "Building Area" : "benefit", 
        "Road Access": "benefit",
        "Distance" : "benefit",
        "Rental Price" : "cost"
    }

    st.subheader("⚖ Bobot Kriteria (Manual)")
    weight_dict = {}
    cols = st.columns(len(criteria))
    for i, c in enumerate(criteria):
        with cols[i]:
            weight_dict[c] = st.number_input(f"Bobot untuk '{c}' dengan rentang 1-5", min_value=0.0, max_value=10.0, value=0.0, step=0.1, key=f"weight_{c}")
    
    types = np.array([1 if criteria_default_type.get(c,"benefit") == "benefit" else 0 for c in criteria])

    
    weights = np.array([weight_dict[c] for c in criteria])
    weights /= weights.sum()
    st.write("🎯 Bobot Kriteria (Ternormalisasi):")
        # Gabungkan ke dalam DataFrame
    summary_df = pd.DataFrame({
        "Kriteria": criteria,
        "Bobot (Ternormalisasi)": weights,
        "Tipe ": types
    })
    
    # Tampilkan
    #st.write("📋 Tabel Kriteria, Bobot, dan Tipe:")
    st.dataframe(summary_df)
    #st.dataframe(pd.Series(weights, index=criteria, name="Bobot", ))

    
else:
    #Profile Matching
    st.subheader("📋 Input Nilai Alternatif terhadap Kriteria")
    empty_data = pd.DataFrame(np.nan, index=alternatives, columns=criteria)
    df = st.data_editor(empty_data, use_container_width=True, key="input_matrix")
    if df.isnull().values.any():
        st.warning("⚠ Harap lengkapi semua nilai pada tabel sebelum menjalankan perhitungan untuk TOPSIS/Profile Matching.")

    st.subheader("🎯 Ideal Profile (untuk Profile Matching)")
    ideal_profile_dict = {}
    cols = st.columns(len(criteria))
    for i, c in enumerate(criteria):
        with cols[i]:
            ideal_profile_dict[c] = st.number_input(f"Ideal '{c}' (1-5)", min_value=0, max_value=5, value=0, key=f"ideal_{c}")
    ideal_profile = pd.Series(ideal_profile_dict, index=criteria)
    st.write("💡 Profil Ideal:")
    st.dataframe(ideal_profile.to_frame(name="Nilai Ideal"))

    st.subheader("⚙ Pengaturan Faktor (untuk Profile Matching)")
    core_factors_options = criteria
    default_core_factors = [criteria[-1]] if criteria else []

    selected_core_factors = st.multiselect(
        "Pilih Kriteria Core Factor (Faktor Inti)",
        options=core_factors_options,
        default=default_core_factors,
        help="Kriteria yang dianggap paling penting. Sisanya akan menjadi Secondary Factor.",
        key="pm_core_factors"
    )

    if not selected_core_factors:
        st.warning("Setidaknya satu Core Factor harus dipilih untuk Profile Matching.")
        st.stop()

    cf_weight = st.number_input(f"Bobot Core Factor", min_value=0.0, max_value=1.0, value=0.0, step=0.05)
    sf_weight = 1 - cf_weight

# Tombol jalankan perhitungan
run_calc = st.button("▶ Jalankan Perhitungan")

if run_calc:
    if method == "AHP":
        if len(alternative_weights_per_criterion) != len(criteria):
            st.error("Perbandingan alternatif belum lengkap untuk semua kriteria. Lengkapi terlebih dahulu.")
        elif weights_ahp_criteria.empty:
            st.error("Bobot kriteria AHP belum terhitung dengan benar.")
        else:
            # Hitung skor akhir AHP per alternatif = sum bobot kriteria * bobot alternatif pada kriteria
            final_scores_ahp = pd.Series(0.0, index=alternatives)
            for c in criteria:
                w_crit = weights_ahp_criteria.get(c, 0)
                w_alts = alternative_weights_per_criterion.get(c, pd.Series(0, index=alternatives))
                final_scores_ahp += w_crit * w_alts

            st.subheader("📈 Hasil Perhitungan AHP")
            st.write(final_scores_ahp.sort_values(ascending=False).to_frame("Skor Akhir"))
            st.write("*Alternatif terbaik berdasarkan AHP:*")
            st.success(final_scores_ahp.idxmax())

    elif method == "TOPSIS":
        if df.isnull().values.any():
            st.error("Isi semua nilai alternatif terlebih dahulu.")
        else:
            data = df.values.astype(float)
            w = weights

            # Normalisasi
            norm_data = data / np.sqrt((data ** 2).sum(axis=0))

            # Bobot
            weighted_data = norm_data * w

            # Tentukan ideal positif dan negatif (max/min)
            # ideal_pos = np.max(weighted_data, axis=0)
            # ideal_neg = np.min(weighted_data, axis=0)
            ideal_pos = np.where(types == 1, weighted_data.max(), weighted_data.min())
            ideal_neg = np.where(types == 1, weighted_data.min(), weighted_data.max())

            # Jarak ke ideal positif dan negatif
            dist_pos = np.sqrt(((weighted_data - ideal_pos) ** 2).sum(axis=1))
            dist_neg = np.sqrt(((weighted_data - ideal_neg) ** 2).sum(axis=1))

            # Skor preferensi
            scores_topsis = dist_neg / (dist_pos + dist_neg)

            topsis_result = pd.Series(scores_topsis, index=alternatives).sort_values(ascending=False)

            st.subheader("📈 Hasil Perhitungan TOPSIS")
            st.write(topsis_result.to_frame("Skor"))
            st.write("*Alternatif terbaik berdasarkan TOPSIS:*")
            st.success(topsis_result.idxmax())

    elif method == "Profile Matching":
        if df.isnull().values.any():
            st.error("Isi semua nilai alternatif terlebih dahulu.")
        else:
            # Scale input values to 1-5 range first
            def scale_to_five(val):
                if val >= 90: return 5
                elif val >= 80: return 4
                elif val >= 70: return 3
                elif val >= 60: return 2
                elif val >= 50: return 1
                else: return 1
                
            df_scaled = df.applymap(scale_to_five)
            st.write("📏 Data Skala 1–5", df_scaled)

            # Get core factor indices
            cf_indices = [criteria.index(c) for c in selected_core_factors]
            
            # Run profile matching with the improved function
            gap_weights = {
                0: 5, 1: 4.5, -1: 4.5, 2: 4, -2: 4,
                3: 3.5, -3: 3.5, 4: 3, -4: 3, 5: 2.5, -5: 2.5
            }
            
            df_gap = df_scaled - ideal_profile
            df_wgap = df_gap.applymap(lambda x: gap_weights.get(int(x), 0))
            
            cf_cols = df_scaled.columns[cf_indices]
            sf_cols = df_scaled.columns.drop(cf_cols)
            
            ncf = df_wgap[cf_cols].mean(axis=1)
            nsf = df_wgap[sf_cols].mean(axis=1) if len(sf_cols) > 0 else 0

            final_scores_pm = cf_weight * ncf + sf_weight * nsf

            st.subheader("📈 Hasil Perhitungan Profile Matching")
            st.write(final_scores_pm.sort_values(ascending=False).to_frame("Skor"))
            st.write("Alternatif terbaik berdasarkan Profile Matching:")
            st.success(final_scores_pm.idxmax())


            # data = df.astype(float)
            # ideal_profile = pd.Series(ideal_profile_dict)

            # core_factors = selected_core_factors
            # secondary_factors = [c for c in criteria if c not in core_factors]

            # # Hitung gap untuk tiap alternatif dan kriteria
            # gaps = data.subtract(ideal_profile, axis=1)

            # # Ubah gap menjadi nilai bobot sesuai dengan skala gap (bisa disesuaikan)
            # def gap_weight(g):
            #     if g == 0:
            #         return 5
            #     elif g == -1:
            #         return 4.5
            #     elif g == 1:
            #         return 4
            #     elif g == -2:
            #         return 3.5
            #     elif g == 2:
            #         return 3
            #     elif g == -3:
            #         return 2.5
            #     elif g == 3:
            #         return 2
            #     elif g == -4:
            #         return 1.5
            #     elif g == 4:
            #         return 1
            #     else:
            #         return 1

            # weight_gaps = gaps.applymap(gap_weight)

            # # Hitung core factor dan secondary factor scores per alternatif
            # cf_scores = weight_gaps[core_factors].mean(axis=1)
            # sf_scores = weight_gaps[secondary_factors].mean(axis=1) if secondary_factors else 0

            # final_scores_pm = cf_weight * cf_scores + sf_weight * sf_scores

            # st.subheader("📈 Hasil Perhitungan Profile Matching")
            # st.write(final_scores_pm.sort_values(ascending=False).to_frame("Skor"))
            # st.write("*Alternatif terbaik berdasarkan Profile Matching:*")
            # st.success(final_scores_pm.idxmax())