File size: 3,364 Bytes
e92dcbe |
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 |
import os
import numpy as np
import pandas as pd
from copy import copy
import joblib
import tensorflow as tf
from mastml.models import EnsembleModel
import keras
from keras.models import Sequential
from keras.layers import Dense, Dropout, BatchNormalization
from keras.wrappers.scikit_learn import KerasRegressor
import matplotlib.pyplot as plt
from scipy.spatial.distance import cdist
def rebuild_model(model_folder):
PATH = os.getcwd()
# We need to define the function that builds the network architecture
def keras_model():
model = Sequential()
model.add(Dense(1024, input_dim=9, kernel_initializer='normal', activation='relu'))
model.add(Dropout(0.3))
model.add(Dense(1024, kernel_initializer='normal', activation='relu'))
model.add(Dropout(0.3))
model.add(Dense(1, kernel_initializer='normal'))
model.compile(loss='mean_squared_error', optimizer='adam')
return model
model_keras = KerasRegressor(build_fn=keras_model, epochs=250, batch_size=100, verbose=0)
model_bagged_keras_rebuild = EnsembleModel(model=model_keras, n_estimators=10)
num_models = 10
models = list()
for i in range(num_models):
models.append(tf.keras.models.load_model(os.path.join(PATH, 'RPV_model'+'/'+model_folder+'/keras_model_' + str(i))))
model_bagged_keras_rebuild.model.estimators_ = models
model_bagged_keras_rebuild.model.estimators_features_ = [np.arange(0, 9) for i in models]
return model_bagged_keras_rebuild
def get_preds_ebars(model, df_featurized, preprocessor, return_ebars=True):
preds_each = list()
ebars_each = list()
df_featurized_scaled = preprocessor.transform(pd.DataFrame(df_featurized))
if return_ebars == True:
for i, x in df_featurized_scaled.iterrows():
preds_per_data = list()
for m in model.model.estimators_:
preds_per_data.append(m.predict(pd.DataFrame(x).T, verbose=0)) #pd.DataFrame(x).T
preds_each.append(np.mean(preds_per_data))
ebars_each.append(np.std(preds_per_data))
else:
preds_each = model.predict(df_featurized_scaled)
ebars_each = [np.nan for i in range(preds_each.shape[0])]
if return_ebars == True:
a = -0.041
b = 2.041
c = 3.124
ebars_each_recal = a*np.array(ebars_each)**2 + b*np.array(ebars_each) + c
else:
ebars_each_recal = ebars_each
return np.array(preds_each).ravel(), np.array(ebars_each_recal).ravel()
def make_predictions_DNN(df_featurized, model_folder):
PATH = os.getcwd()
# Rebuild the saved model
model = rebuild_model(model_folder)
# Normalize the input features
preprocessor = joblib.load(os.path.join(PATH, os.path.join('RPV_model', model_folder+'/StandardScaler.pkl')))
# Get predictions and error bars from model
preds, ebars = get_preds_ebars(model, df_featurized, preprocessor, return_ebars=True)
pred_dict = {'preds':preds,
'ebars':ebars}
return pd.DataFrame(pred_dict)
def test(df):
#pred_dict = {'preds': ['This is a test'], 'ebars': ['This is a test']}
pred_arr = np.array([['here is some data'], ['here are some ebars']])
model = keras.models.load_model('keras_model_0')
return pred_arr
#return pd.DataFrame(pred_dict)
#return np.sqrt(x)
|