project_text_classification/plot.py at master · Eagleseb/project_text_classification · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
import matplotlib.pyplot as plt
import numpy as np
from sklearn import ensemble
from sklearn.decomposition import PCA
from sklearn.model_selection import cross_val_score


from loader import load_data, prepare_data, get_frequencies


def pca(X_train, n_components):
    pca = PCA(n_components=n_components)
    pca.fit(X_train)

    distrib = np.r_[0, np.cumsum(pca.explained_variance_ratio_)]

    x = np.arange(0, n_components + 1)
    plt.bar(x, distrib)
    plt.axhline(y=.95, xmin=0, xmax=n_components, label="95% threshold", c="r")
    plt.axvline(x=(n_components+1)**.95 - 1, ymin=0, ymax=1, c="r")
    plt.plot(x, np.log(x + 1) / np.log(n_components + 1), label="log(x + 1)/log(n_components + 1)", c="black")
    plt.legend()
    plt.title("Cumulative explained variance of\nthe {} principal components out of {} features"
              .format(n_components, X_train.shape[1]))
    plt.savefig("output/explained_variance.eps")


def n_estimators_computation(num_estimators, X_train, y_train):

    clf = ensemble.RandomForestClassifier(n_estimators=num_estimators)

    clf.fit(X_train, y_train)
    scores = cross_val_score(clf, X_train, y_train, cv=3, n_jobs=-1)

    scores_mean = scores.mean() * 100
    print("Cross validated score: {:.1f} +/- {:.1f}".format(scores.mean() * 100, scores.std() * 100))

    return scores_mean


def n_estimators_plot(X_train, y_train, lower_b, upper_b, num_interval):
    values = np.logspace(np.log10(lower_b), np.log10(upper_b), num=num_interval)
    scores_estimators = np.zeros(len(values))

    for idx, v in enumerate(values) :
        print("Computation for the number of estimators : {}".format(int(np.floor(v))))
        scores_estimators[idx] = n_estimators_computation(int(np.floor(v)), X_train, y_train)

    plt.figure()
    plt.plot(values, scores_estimators, label = "Scores for different values of number of estimators")
    # plt.axhline(y=.95, xmin=0, xmax=n_components, label="95% threshold", c="r")
    # plt.axvline(x=(n_components+1)**.95 - 1, ymin=0, ymax=1, c="r")
    plt.legend()
    plt.title("Performance on the cross-validation for different numbers of estimators with Random Forest Classifier")
    plt.savefig('fig/estimators.eps')

    return scores_estimators


def plot_frequencies(train_pos, train_neg):
    freq_pos = get_frequencies(train_pos)
    freq_neg = get_frequencies(train_neg)

    # freq_pos = sorted(freq_pos.items(), key=lambda kv: kv[1], reverse=True)
    # freq_neg = sorted(freq_neg.items(), key=lambda kv: kv[1], reverse=True)

    diff = {}
    for (k, v) in freq_pos.items():
        diff[k] = v
    for (k, v) in freq_neg.items():
        diff[k] = diff.get(k, 0) - v

    diff = sorted(diff.items(), key=lambda kv: kv[1], reverse=True)
    abs_diff = sorted(diff, key=lambda p: abs(p[1]))

    # first_significant = np.argwhere(list(map(lambda p: abs(p[1]) > 0.00001, abs_diff)))[0, 0]

    x = np.linspace(0, 100, 50)
    plt.figure(figsize=(10, 11))
    plt.barh(x, list(map(lambda p: p[1], diff))[:50], color='g')
    plt.yticks(x, list(map(lambda p: p[0], diff))[:50])
    plt.savefig('output/positive_words.eps')
    plt.figure(figsize=(10, 11))
    plt.barh(x, list(map(lambda p: p[1], diff))[-50:], color='r')
    plt.yticks(x, list(map(lambda p: p[0], diff))[-50:])
    plt.savefig('output/negative_words.eps')
    plt.figure()
    plt.plot(np.arange(len(abs_diff)), list(map(lambda p: abs(p[1]), abs_diff)), color='orange')
    plt.semilogy()
    plt.savefig('output/freq_diff.eps')


def main():
    np.random.seed(42)

    # print("Loading data")
    # X_train, y_train = load_data('data/glove.twitter.27B/glove.twitter.27B.200d.txt',
    #                             'data/train_pos.txt', 'data/train_neg.txt')
    #
    # print("Plotting explained variance")
    # pca(X_train, X_train.shape[1])
    #
    # print("Preparing data")
    # X_train, y_train = prepare_data(X_train, y_train)

    # print("Plotting the performance depending on the number of estimators")
    # n_estimators_plot(X_train, y_train, 1, 50, 10)

    plot_frequencies('data/train_pos.txt', 'data/train_neg.txt')


if __name__ == '__main__':
    main()