PredictiveMonitoringWithText/SequenceEncoder.py at master · irhete/PredictiveMonitoringWithText · 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
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
import pandas as pd
from sklearn.feature_extraction import DictVectorizer as DV

class SequenceEncoder():

    def __init__(self, nr_events, event_nr_col, case_id_col, label_col, static_cols=[], dynamic_cols=[],
                 last_state_cols=[], cat_cols=[], oversample_fit=True, minority_label="positive", fillna=True,
                 random_state=None, max_events=200, dyn_event_marker="dynevent", last_event_marker="lastevent",
                case_length_col = "case_length", pre_encoded=False):
        self.nr_events = nr_events
        self.static_cols = static_cols
        self.dynamic_cols = dynamic_cols
        self.last_state_cols = last_state_cols
        self.cat_cols = cat_cols
        self.event_nr_col = event_nr_col
        self.case_id_col = case_id_col
        self.label_col = label_col
        self.oversample_fit = oversample_fit
        self.minority_label = minority_label
        self.random_state = random_state
        self.fillna = fillna
        self.dyn_event_marker = dyn_event_marker
        self.last_event_marker = last_event_marker
        self.max_events = max_events
        self.case_length_col = case_length_col
        self.pre_encoded = pre_encoded

        self.fitted_columns = None


    def fit(self, X):
        return self

    def fit_transform(self, X):
        data = self._encode(X)
        if self.oversample_fit:
            data = self._oversample(data)
        return data

    def transform(self, X):
        data = self._encode(X)
        return data

    def pre_encode(self, X):
        # encode static cols
        if self.label_col not in self.static_cols:
            self.static_cols.append(self.label_col)
        if self.case_id_col not in self.static_cols:
            self.static_cols.append(self.case_id_col)
        data_final = X[X[self.event_nr_col] == 1][self.static_cols]

        # encode dynamic cols
        for i in range(1, self.max_events+1):
            data_selected = X[X[self.event_nr_col] == i][[self.case_id_col] + self.dynamic_cols]
            data_selected.columns = [self.case_id_col] + ["%s_%s%s"%(col, self.dyn_event_marker, i) for col in self.dynamic_cols]
            data_final = pd.merge(data_final, data_selected, on=self.case_id_col, how="left")


        # encode last state cols
        for i in range(1, self.max_events+1):
            data_selected = X[X[self.event_nr_col] == i][[self.case_id_col] + self.last_state_cols]
            data_selected.columns = [self.case_id_col] + ["%s_%s%s"%(col, self.last_event_marker, i) for col in self.last_state_cols]
            data_final = pd.merge(data_final, data_selected, on=self.case_id_col, how="left")
            if i > 1:
                for col in self.last_state_cols:
                    missing = pd.isnull(data_final["%s_%s%s"%(col, self.last_event_marker, i)])
                    data_final["%s_%s%s"%(col, self.last_event_marker, i)].loc[missing] = data_final["%s_%s%s"%(col, self.last_event_marker, i-1)].loc[missing]


        # make categorical
        dynamic_cat_cols = [col for col in self.cat_cols if col in self.dynamic_cols]
        static_cat_cols = [col for col in self.cat_cols if col in self.static_cols]
        categorical_cols = ["%s_%s%s"%(col, self.dyn_event_marker, i) for i in range(1, self.max_events+1) for col in dynamic_cat_cols] + static_cat_cols
        cat_df = data_final[categorical_cols]
        cat_dict = cat_df.T.to_dict().values()
        vectorizer = DV( sparse = False )
        vec_cat_dict = vectorizer.fit_transform(cat_dict)
        cat_data = pd.DataFrame(vec_cat_dict, columns=vectorizer.feature_names_)
        data_final = pd.concat([data_final.drop(categorical_cols, axis=1), cat_data], axis=1)

        data_final = pd.merge(data_final, X.groupby(self.case_id_col)[self.event_nr_col].agg({"case_length": "max"}).reset_index(), on=self.case_id_col, how="left")

        # fill NA
        if self.fillna:
            for col in data_final:
                dt = data_final[col].dtype
                if dt == int or dt == float:
                    data_final[col].fillna(0, inplace=True)
                else:
                    data_final[col].fillna("", inplace=True)

        return data_final


    def _encode(self, X):
        if self.pre_encoded:
            rel_cols = X.columns[~X.columns.str.contains('|'.join(["%s%s"%(self.dyn_event_marker, k) for k in
                                                                       range(self.nr_events+1, self.max_events+1)] +
                                                                  [self.last_event_marker]))]
            rel_cols = rel_cols | X.columns[X.columns.str.endswith("%s%s"%(self.last_event_marker, self.nr_events))]
            selected = X[rel_cols]
            selected = selected[selected[self.case_length_col] >= self.nr_events]
            return selected.drop(self.case_length_col, axis=1)
        else:
            return self._complex_encode(X)

    def _complex_encode(self, X):
        # encode static cols
        if self.label_col not in self.static_cols:
            self.static_cols.append(self.label_col)
        if self.case_id_col not in self.static_cols:
            self.static_cols.append(self.case_id_col)
        data_final = X[X[self.event_nr_col] == 1][self.static_cols]

        # encode dynamic cols
        for i in range(1, self.nr_events+1):
            data_selected = X[X[self.event_nr_col] == i][[self.case_id_col] + self.dynamic_cols]
            data_selected.columns = [self.case_id_col] + ["%s_%s"%(col, i) for col in self.dynamic_cols]
            data_final = pd.merge(data_final, data_selected, on=self.case_id_col, how="right")

        # encode last state cols
        for col in self.last_state_cols:
            data_final = pd.merge(data_final, X[X[self.event_nr_col] == self.nr_events][[self.case_id_col, col]], on=self.case_id_col, how="right")
            for idx, row in data_final.iterrows():
                current_nr_events = self.nr_events - 1
                while pd.isnull(data_final.loc[idx, col]) and current_nr_events > 0:
                    data_final.loc[idx, col] = X[(X[self.case_id_col] == row[self.case_id_col]) & (X[self.event_nr_col] == current_nr_events)].iloc[0][col]
                    current_nr_events -= 1


        # make categorical
        dynamic_cat_cols = [col for col in self.cat_cols if col in self.dynamic_cols]
        static_cat_cols = [col for col in self.cat_cols if col in self.static_cols]
        catecorical_cols = ["%s_%s"%(col, i) for i in range(1, self.nr_events+1) for col in dynamic_cat_cols] + static_cat_cols
        cat_df = data_final[catecorical_cols]
        cat_dict = cat_df.T.to_dict().values()
        vectorizer = DV( sparse = False )
        vec_cat_dict = vectorizer.fit_transform(cat_dict)
        cat_data = pd.DataFrame(vec_cat_dict, columns=vectorizer.feature_names_)
        data_final = pd.concat([data_final.drop(catecorical_cols, axis=1), cat_data], axis=1)

        if self.fitted_columns is not None:
            missing_cols = self.fitted_columns[~self.fitted_columns.isin(data_final.columns)]
            for col in missing_cols:
                data_final[col] = 0
            data_final = data_final[self.fitted_columns]
        else:
            self.fitted_columns = data_final.columns

        # fill NA
        if self.fillna:
            for col in data_final:
                dt = data_final[col].dtype
                if dt == int or dt == float:
                    data_final[col].fillna(0, inplace=True)
                else:
                    data_final[col].fillna("", inplace=True)

        return data_final

    def _oversample(self, X):
        oversample_count = sum(X[self.label_col] != self.minority_label) - sum(X[self.label_col] == self.minority_label)

        if oversample_count > 0 and sum(X[self.label_col] == self.minority_label) > 0:
            oversampled_data = X[X[self.label_col]==self.minority_label].sample(oversample_count, replace=True, random_state=self.random_state)
            X = pd.concat([X, oversampled_data])

        return X