-
Notifications
You must be signed in to change notification settings - Fork 0
/
model_predict.py
48 lines (42 loc) · 1.49 KB
/
model_predict.py
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
"""
makes prediction on one set of data with one model
"""
from tensorflow import keras
import pandas as pd
import numpy as np
def make_prediction(
model_id,
x_data,
time_stamps,
threshold,
values,
model_path,
anomaly_type="anomaly",
manual_anomaly=None,
):
"""
model_id: string, name of the model to load
x_data: np array of shape (num_rows, time_steps, 1)
time_stamps: time_stamps: list, timestamps of prediction points,
find this out before passing
threshold: float, for flagging as anomalous
values: unscaled values to be added back to the resulting dataframe
model_path: string, path to saved model
anomaly_type: string, what to name the
anomaly column (eg manual_anomaly, model_anomaly, or realtime_anomaly)
"""
# make predictions
model = keras.models.load_model(model_path + model_id)
x_test_pred = model.predict(x_data, verbose=0)
# format predictions
test_mae_loss = np.mean(np.abs(x_test_pred - x_data), axis=1)
test_score_df = pd.DataFrame(time_stamps, columns=["DateTime"])
test_score_df["loss"] = test_mae_loss
test_score_df["threshold"] = threshold
test_score_df[anomaly_type] = test_score_df["loss"] > test_score_df["threshold"]
test_score_df["uniqueID"] = model_id
test_score_df["val_num"] = values
if manual_anomaly is not None:
test_score_df["manual_anomaly"] = manual_anomaly
test_score_df.set_index("DateTime", drop=True, inplace=True)
return test_score_df