import sys
sys.path.insert(0, "..")

import math
from collections import Counter

import numpy as np
import pandas as pd
import networkx as nx
from sklearn.cluster import DBSCAN, KMeans
from sklearn.metrics import confusion_matrix, f1_score, precision_score, recall_score, silhouette_score
from mlxtend.preprocessing import TransactionEncoder
from mlxtend.frequent_patterns import apriori, association_rules

from common.datasets import play_tennis, transaksi_contoh, dokumen_teks

def cek(nama, saya, benar, tol=1e-3):
    if saya is None:
        print(f"{nama}: jawaban benar = {benar}")
    else:
        ok = abs(saya - benar) <= tol
        print(f"{nama}: {'✅ BENAR' if ok else '❌ SALAH'} (kamu {saya}, kunci {round(benar, 4)})")

transaksi = transaksi_contoh()
for i, t in enumerate(transaksi, 1):
    print(f"T{i}: {t}")

jawaban_saya = {"support": None, "confidence": None, "lift": None}

te = TransactionEncoder()
X = pd.DataFrame(te.fit_transform(transaksi), columns=te.columns_)
fi = apriori(X, min_support=0.2, use_colnames=True)
rules = association_rules(fi, metric="confidence", min_threshold=0.0)
r = rules[(rules.antecedents == frozenset({"I1", "I2"})) & (rules.consequents == frozenset({"I5"}))].iloc[0]
for k in ["support", "confidence", "lift"]:
    cek(k, jawaban_saya[k], r[k])

df = play_tennis()
df

jawaban_saya = {"entropy": None, "outlook": None, "temperature": None, "humidity": None, "wind": None}

def entropy(s):
    p = s.value_counts(normalize=True)
    return -(p * np.log2(p)).sum()

def gain(df, a, target="play"):
    return entropy(df[target]) - sum(len(g) / len(df) * entropy(g[target]) for _, g in df.groupby(a))

cek("entropy", jawaban_saya["entropy"], entropy(df["play"]))
for a in ["outlook", "temperature", "humidity", "wind"]:
    cek(f"gain {a}", jawaban_saya[a], gain(df, a))

jawaban_saya = {"accuracy": None, "precision": None, "recall": None, "f1": None}

y = np.array([1] * 40 + [0] * 10 + [1] * 20 + [0] * 30)
p = np.array([1] * 40 + [1] * 10 + [0] * 20 + [0] * 30)
print(confusion_matrix(y, p, labels=[1, 0]))
cek("accuracy", jawaban_saya["accuracy"], (y == p).mean())
cek("precision", jawaban_saya["precision"], precision_score(y, p))
cek("recall", jawaban_saya["recall"], recall_score(y, p))
cek("f1", jawaban_saya["f1"], f1_score(y, p))

P_ = np.array([[1, 1], [1.5, 2], [4, 4], [5, 7], [3.5, 5], [4.5, 5], [3.5, 4.5]])
km = KMeans(2, init=np.array([[1, 1], [5, 7.0]]), n_init=1, max_iter=1).fit(P_)
print("label   :", km.labels_)
print("centroid:\n", km.cluster_centers_)
print("silhouette:", round(silhouette_score(P_, km.labels_), 3))

Q = np.array([[0, 0], [0, 1], [1, 0], [1, 1], [2.5, 1], [10, 10]])
db = DBSCAN(eps=1.5, min_samples=4).fit(Q)
jenis = ["noise" if l == -1 else ("core" if i in db.core_sample_indices_ else "border")
         for i, l in enumerate(db.labels_)]
pd.DataFrame({"titik": list(map(tuple, Q)), "jenis": jenis})

jawaban_saya = {"blok": None, "total_mb": None}
cek("jumlah blok", jawaban_saya["blok"], math.ceil(1000 / 128), tol=0)
cek("total MB", jawaban_saya["total_mb"], 1000 * 3, tol=0)

peta = [(w, 1) for d in dokumen_teks() for w in d.split()]          # map
grup = {}
for w, v in peta:                                                   # shuffle
    grup.setdefault(w, []).append(v)
hasil = {w: sum(v) for w, v in grup.items()}                        # reduce
print("frekuensi 'data':", hasil["data"], "| cek Counter:", Counter(w for d in dokumen_teks() for w in d.split())["data"])

jawaban_saya = {"S3": None, "p_lsh": None, "rA": None, "rB": None, "rC": None}
s = 0
for x in [10, 10, 10]:
    s = 0.9 * s + x
cek("S3", jawaban_saya["S3"], s)
cek("P(kandidat)", jawaban_saya["p_lsh"], 1 - (1 - 0.8 ** 5) ** 20)

r0 = np.ones(3) / 3
M = np.array([[0, 0, 1], [0.5, 0, 0], [0.5, 1, 0]])  # kolom = sumber A, B, C
r1 = 0.15 / 3 + 0.85 * M @ r0
for nama, val in zip(["rA", "rB", "rC"], r1):
    cek(nama, jawaban_saya[nama], val)
print("PageRank konvergen (networkx):", {k: round(v, 4) for k, v in nx.pagerank(nx.DiGraph([("A","B"),("A","C"),("B","C"),("C","A")])).items()})

jawaban_saya = {"c": None, "s3": None, "s12": None}
c = 2 * (math.log(255) + 0.5772156649) - 2 * 255 / 256
cek("c(256)", jawaban_saya["c"], c)
cek("s(E=3)", jawaban_saya["s3"], 2 ** (-3 / c))
cek("s(E=12)", jawaban_saya["s12"], 2 ** (-12 / c))