util.py

import itertools
import subprocess
import os
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import torch
import copy
import torch.nn.functional as F
import random
import csv
import sys
from torch import nn
from tqdm import tqdm_notebook, trange, tqdm
from pytorch_pretrained_bert.optimization import BertAdam
from pytorch_pretrained_bert.modeling import WEIGHTS_NAME,CONFIG_NAME,BertPreTrainedModel,BertModel
from pytorch_pretrained_bert.tokenization import BertTokenizer
from torch.utils.data import (DataLoader, RandomSampler, SequentialSampler, TensorDataset)
from datetime import datetime
from sklearn.cluster import KMeans
from sklearn.metrics import confusion_matrix,normalized_mutual_info_score, adjusted_rand_score, accuracy_score
from scipy.optimize import linear_sum_assignment
from sklearn import metrics

def hungray_aligment(y_true, y_pred):
    D = max(y_pred.max(), y_true.max()) + 1
    w = np.zeros((D, D))
    for i in range(y_pred.size):
        w[y_pred[i], y_true[i]] += 1

    ind = np.transpose(np.asarray(linear_sum_assignment(w.max() - w)))
    return ind, w

def clustering_accuracy_score(y_true, y_pred):
    ind, w = hungray_aligment(y_true, y_pred)
    acc = sum([w[i, j] for i, j in ind]) / y_pred.size
    return acc

def clustering_score(y_true, y_pred):
    return {'ACC': round(clustering_accuracy_score(y_true, y_pred)*100, 2),
            'ARI': round(adjusted_rand_score(y_true, y_pred)*100, 2),
            'NMI': round(normalized_mutual_info_score(y_true, y_pred)*100, 2)}