lazyfca15/test_implication.py at master · fdrstrok/lazyfca15 · 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
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
import sys
import random


# attrib_names = [ 'class','a1','a2','a3','a4','a5','a6' ]
attrib_names = [
    'top-left-square',
    'top-middle-square',
    'top-right-square',
    'middle-left-square',
    'middle-middle-square',
    'middle-right-square',
    'bottom-left-square',
    'bottom-middle-square',
    'bottom-right-square',
    'class'
]


def make_intent(example):
    global attrib_names
    return set([i+':'+str(k) for i, k in zip(attrib_names, example)])


cv_res = {
    "positive_positive": 0,
    "positive_negative": 0,
    "negative_positive": 0,
    "negative_negative": 0,
    "contradictory": 0,
    "total": 0,
}


def check_intersect(context_plus, context_minus, example, num_sub=1):
    global cv_res
    pos = 0
    neg = 0
    intent = make_intent(example)
    for i in xrange(num_sub):
        t = set(random.sample(example, random.randrange(len(intent))))
        for j in context_plus:
            if t.issubset(j):
                pos += len(t)
        for k in context_minus:
            if t.issubset(k):
                neg += len(t)

    def score(pos, neg):
        return pos * 1. / (neg + 1)

    threshold = 1.1
    if score(pos, neg) > threshold:
        if example[-1] == 'positive':
            cv_res['positive_positive'] += 1
        else:
            cv_res['negative_positive'] += 1
    elif score(neg, pos) > threshold:
        if example[-1] == 'positive':
            cv_res['positive_negative'] += 1
        else:
            cv_res['negative_negative'] += 1
    else:
        cv_res['contradictory'] += 1


def check_hypothesis(context_plus, context_minus, example):
    global cv_res
    eintent = make_intent(example)
    big_context = context_plus + context_minus
    labels = {}
    for e in big_context:
        ei = make_intent(e)
        candidate_intent = ei & eintent
        if not candidate_intent:
            continue
        # print 'candidate_intent'
        # print candidate_intent
        closure = [make_intent(i) for i in big_context
                   if make_intent(i).issuperset(candidate_intent)]
#        print 'closure:'
#        print closure
        res = reduce(lambda x, y: x & y if x & y else x | y, closure)
        # print 'reduced:'
        # print res
        for cs in ['positive', 'negative']:
            if 'class:' + cs in res:
                labels[cs] = True
                labels[cs + '_res'] = candidate_intent

                # print 'classified as %s, reason:' % cs
                # print candidate_intent
                # print res
#    print labels
    if labels.get("positive", False) and labels.get("negative", False):
        cv_res["contradictory"] += 1
        return
    if example[-1] == "positive" and labels.get("positive", False):
        cv_res["positive_positive"] += 1
    if example[-1] == "negative" and labels.get("positive", False):
        cv_res["negative_positive"] += 1
    if example[-1] == "positive" and labels.get("negative", False):
        cv_res["positive_negative"] += 1
    if example[-1] == "negative" and labels.get("negative", False):
        cv_res["negative_negative"] += 1

    '''closure = []
    for candidate in context:
        cintent = make_intent(candidate)
        closure.append(cintent&eintent)
        if cintent.issubset(eintent):
            print "classified!"
            print cintent
            print eintent
    '''
# sanity check:
# check_hypothesis(plus_examples, minus_examples, plus_examples[3])
max_index = sys.argv[1]
for index in xrange(1, int(max_index)):
    index = str(index)
    q = open("train" + index + ".csv", "r")
    train = [a.strip().split(",") for a in q]
    plus = [a for a in train if a[-1] == "positive"]
    minus = [a for a in train if a[-1] == "negative"]

    # print t
    q.close()
    w = open("test" + index + ".csv", "r")
    unknown = [a.strip().split(",") for a in w]
    w.close()

    for elem in unknown:
        cv_res['total'] += 1
        #   print elem
        # print "done"
        # check_hypothesis(plus, minus, elem)
        # check_intersect(plus, minus, elem, len(elem) / 2)
        check_intersect(plus, minus, elem, len(elem) / 2)
    print "done: %s" % index

print cv_res
for k, v in cv_res.iteritems():
    cv_res[k] = v * 1. / cv_res["total"]

print cv_res