3 examples of 'pandas groupby weighted average' in Python

236
def mean(self):
237
    # TODO, there is a lot of copy-paste with the code above
238
    # TODO, we should probably define groupby.aggregate
239
    func = _accumulate_groupby_mean
240
    start = (0, 0)
241
    if isinstance(self.grouper, Streaming):
242
        func = partial(func, index=self.index)
243
        example = self.root.example.groupby(self.grouper.example)
244
        if self.index is not None:
245
            example = example[self.index]
246
        example = example.mean()
247
        stream = self.root.stream.zip(self.grouper.stream)
248
        stream = stream.accumulate(func, start=start, returns_state=True)
249
    else:
250
        func = partial(func, grouper=self.grouper, index=self.index)
251
        example = self.root.example.groupby(self.grouper)
252
        if self.index is not None:
253
            example = example[self.index]
254
        example = example.mean()
255
        stream = self.root.stream.accumulate(func, start=start,
256
                                             returns_state=True)
257
    if isinstance(example, pd.DataFrame):
258
        return StreamingDataFrame(stream, example)
259
    else:
260
        return StreamingSeries(stream, example)

202
def aggregate(group_reports):
203
    summary = {}
204
    top_model = max(group_reports, key=lambda r: r['Dev First Sentence'])
205
    summary['Avg First Sentence'] = np.mean([r['First Sentence'] for r in group_reports])
206
    summary['Std First Sentence'] = np.std([r['First Sentence'] for r in group_reports])
207
    summary['Avg Full Question'] = np.mean([r['Full Question'] for r in group_reports])
208
    summary['Std Full Question'] = np.std([r['Full Question'] for r in group_reports])
209
    summary['First Sentence'] = top_model['First Sentence']
210
    summary['Full Question'] = top_model['Full Question']
211
    summary['Dev First Sentence'] = top_model['Dev First Sentence']
212
    summary['Dev Full Question'] = top_model['Dev Full Question']
213
    summary['first_df'] = top_model['first_df']
214
    summary['full_df'] = top_model['full_df']
215
    summary['char_df'] = top_model['char_df']
216
    summary['fold'] = top_model['fold']
217
    summary['guesser_name'] = top_model['guesser_name']
218
    summary['random_seed'] = top_model['random_seed']
219
    summary['wiki'] = top_model['wiki']
220
    summary['training_time'] = top_model['training_time']
221
222
    stable_scores = []
223
    eager_scores = []
224
    for _, group in top_model['char_df'].sort_values('score', ascending=False).groupby('qanta_id'):
225
        group = group.groupby(['char_index']).first().reset_index()
226
        stable, eager = compute_curve_score(group)
227
        stable_scores.append(stable)
228
        eager_scores.append(eager)
229
    summary['curve_score_stable'] = np.mean(stable_scores)
230
    summary['curve_score_eager'] = np.mean(eager_scores)
231
    return summary

59
def average(average_window, data):
60
    window = []
61
    newdata = []
62
    for v in data:
63
        window.append(v)
64
        if len(window) == average_window:
65
            newdata.append(sum(window)/average_window)
66
            del window[0]
67
    return newdata