import pandas as pd
from dateutil.parser import parse
import matplotlib.pyplot as plt
def TienXuLy():
    # Tiền xử lý
    data = pd.read_csv("/home/huyphuong99/Documents/data_set.csv")
    data = data[data["Customer ID"].notnull()]
    data = data.drop(index = data[data['Invoice'].astype(str).str[0] == 'C'].index)

    # chuan hoa sang ngay
    date_time = []
    X = data.values  # mảng numpy X lưu gía trị của file data.csv
    for i in range(len(X[:, 4])):
        date = parse(X[i, 4])
        date_time.append(date)

    # Tim ngay max
    max_date = date_time[0]
    for i in range(len(date_time)):
        if (max_date < date_time[i]):
            max_date = date_time[i]

    for i in range(len(date_time)):
        date_time[i] = max_date - date_time[i]

    data["Inv_Day"] = date_time
    data["Inv_Day"] = data["Inv_Day"].astype("timedelta64[D]")
    data["Total_Bill"] = data.Quantity * data.Price
    RFM = data.groupby("Customer ID", as_index = False).agg({"Inv_Day": "min", "Invoice": "count", "Total_Bill": "sum"})
    RFM.rename(columns = {"Inv_Day": "Recency", "Invoice": "Frequency", "Total_Bill": "Mometary"}, inplace = True)

    R = RFM["Recency"].tolist()
    for i in range(len(R)):
        R[i] = -R[i]

    F = RFM["Frequency"].tolist()
    m = RFM["Mometary"].tolist()
    #0->100
    M = []
    for i in range(len(m)):
        t = (m[i] - min(m)) / (max(m) - min(m)) * (100 - 0) + 0
        M.append(t)
    ID = RFM["Customer ID"].tolist()

    X = []
    for i in range(len(ID)):
        X.append([R[i],F[i],M[i]])
    return X


import random
import math
import numpy as np

def eucldist(p0, p1):
    dist = 0.0
    for i in range(0, len(p0)):
        dist += (p0[i] - p1[i]) ** 2
    return math.sqrt(dist)

def kmeans(k, datapoints):

    d = len(datapoints[0])
    i = 0
    cluster = [0] * len(datapoints)
    prev_cluster = [-1] * len(datapoints)

    cluster_centers = []
    for i in range(0, k):

         cluster_centers += [random.choice(datapoints)]

         force_recalculation = False
    while (cluster != prev_cluster) or (force_recalculation):

        prev_cluster = list(cluster)
        force_recalculation = False
        i += 1

        for p in range(0, len(datapoints)):
            min_dist = float("inf")

            for c in range(0, len(cluster_centers)):
                dist = eucldist(datapoints[p], cluster_centers[c])
                if (dist < min_dist):
                    min_dist = dist
                    cluster[p] = c
        for k in range(0, len(cluster_centers)):
            new_center = [0] * d
            members = 0
            for p in range(0, len(datapoints)):
                if (cluster[p] == k):
                    for j in range(0, d):
                        new_center[j] += datapoints[p][j]
                    members += 1
            for j in range(0, d):
                if members != 0:
                    new_center[j] = new_center[j] / members
                else:
                    new_center = random.choice(datapoints)
                    force_recalculation = True
                    print("Forced Recalculation...")
            cluster_centers[k] = new_center

    #print("======== Results ========")
    #print("Clusters", cluster_centers)
    #print("Iterations", i)
    #print("Assignments", cluster)
    return cluster_centers, cluster


if __name__ == "__main__":
    datapoints = TienXuLy()
    # K - Number of Clusters
    sse = []
    K = range(1,16)
    for k in K:
        lable, cluster = kmeans(k, datapoints)
        W = 0
        for i in range(k):
            w = 0
            for j in range(len(cluster)):
                if i == cluster[j]:
                    dist_ct = eucldist(lable[i],datapoints[j])
                    w += dist_ct
            W += w
        print(k)
        print(W)
        sse.append(W)
    plt.plot(K, sse, 'bx-')
    plt.xlabel('\nValues of K\n')
    plt.ylabel('Distortion')
    plt.title('\nThe Elbow Method using Distortion\n')
    plt.show()
    print("\n")
    '''k =
    labelM, clusterM = kmeans(k,datapoints)
    print(labelM)
    print(clusterM)'''