Skip to content

Commit dc335b4

Browse files
setu4993setu4993
committed
Updated notebook comments
1 parent be3f68c commit dc335b4

File tree

2 files changed

+304
-9
lines changed

2 files changed

+304
-9
lines changed

‎.ipynb_checkpoints/predict_example-checkpoint.ipynb

Lines changed: 265 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,269 @@
11
{
2-
"cells": [],
3-
"metadata": {},
2+
"cells": [
3+
{
4+
"cell_type": "markdown",
5+
"metadata": {},
6+
"source": [
7+
"### Example for predicting using the RNN\n",
8+
"\n",
9+
"This is an example for running the RNN and get water demand predictions.\n",
10+
"\n",
11+
"Requirements:\n",
12+
"- Python 3.6+\n",
13+
"- NumPy 1.14+ (http://www.numpy.org/)\n",
14+
"\n",
15+
"To run this Notebook locally:\n",
16+
"- Jupyter (https://jupyter.org/)\n",
17+
"\n",
18+
"___Note:___ This model is an online prediction model. The current version of code does provides a separate option for training the RNN. Refer to [training_example.ipynb](./training_example.ipynb) for training the RNN."
19+
]
20+
},
21+
{
22+
"cell_type": "markdown",
23+
"metadata": {},
24+
"source": [
25+
"#### Step 1: Import external libraries and the `rnn` package"
26+
]
27+
},
28+
{
29+
"cell_type": "code",
30+
"execution_count": 1,
31+
"metadata": {},
32+
"outputs": [],
33+
"source": [
34+
"import os\n",
35+
"from rnn import *"
36+
]
37+
},
38+
{
39+
"cell_type": "markdown",
40+
"metadata": {},
41+
"source": [
42+
"#### Step 2: Specify the location on disk of the data directory and the input CSV file (optional)"
43+
]
44+
},
45+
{
46+
"cell_type": "code",
47+
"execution_count": 2,
48+
"metadata": {},
49+
"outputs": [],
50+
"source": [
51+
"DATA_LOCATION = os.getcwd()\n",
52+
"INPUT_FILE = os.path.join(DATA_LOCATION, 'input_files', 'water_6_17_reduced_2_cap_rs.csv')"
53+
]
54+
},
55+
{
56+
"cell_type": "markdown",
57+
"metadata": {},
58+
"source": [
59+
"#### Step 3: To initialize the predictor RNN, create a `rnn_predict` object.\n",
60+
"\n",
61+
"Required parameters:\n",
62+
"- `weights`: Location on disk of the weights file. Typically a Python serialized object (Pickle) file.\n",
63+
"\n",
64+
"Optional parameters:\n",
65+
"- `alpha`: Learning rate for the network, typically a value in the range `(0, 1)`. Default value: `0.38`.\n",
66+
"- `log`: Boolean value specifying if the execution should be logged to the console. Recommended if running from the command line or in an IDE. Default value: `False`.\n",
67+
"- `h_bias`: Boolean value specifying if hidden layer should have a bias node. Currently trained models include a hidden layer bias. Default value: `True`."
68+
]
69+
},
70+
{
71+
"cell_type": "code",
72+
"execution_count": 3,
73+
"metadata": {},
74+
"outputs": [],
75+
"source": [
76+
"predictor = rnn_predict(os.path.join(DATA_LOCATION, 'input_files', 'daily_rnn_retrained_16.pickle'))"
77+
]
78+
},
79+
{
80+
"cell_type": "markdown",
81+
"metadata": {},
82+
"source": [
83+
"At this point, the model is ready to make predictions.\n",
84+
"\n",
85+
"#### Step 4a: Give the network a list of inputs to make predictions\n",
86+
"\n",
87+
"Required parameters:\n",
88+
"- `inputs`: A list of `[day_of_the_year, maximum_temperature, precipitation]` lists. These should be real world observed or predicted values.\n",
89+
"\n",
90+
"Optional parameters:\n",
91+
"- `networks`: A list of the same length as `inputs` list with values of the range `[0, 11]`, specifying what network each prediction to run on, typically `num_month - 1`. If not specified, all predictions will be made for the month of January.\n",
92+
"- `targets`: A list of the same length as `inputs` of the observed values for the day. The target values are required for updating the weights of the model.\n",
93+
"- `recal`: Boolean value specifying if the network weights should be recalculated. Recommended if using observed demand as target, to improve future predictions. If set to `True`, requires `targets` to be specified. Default value: `False`.\n",
94+
"\n",
95+
"Output:\n",
96+
"- `predictions`: A list of the same length as `inputs` of the predictions made by the network."
97+
]
98+
},
99+
{
100+
"cell_type": "code",
101+
"execution_count": 4,
102+
"metadata": {
103+
"scrolled": false
104+
},
105+
"outputs": [
106+
{
107+
"name": "stdout",
108+
"output_type": "stream",
109+
"text": [
110+
"The prediction for inputs [1, 71, 0] for network 0 is 110.72\n",
111+
"The prediction for inputs [2, 101, 121] for network 0 is 110.71\n"
112+
]
113+
}
114+
],
115+
"source": [
116+
"inputs = [[1, 71, 0], [2, 101, 121]]\n",
117+
"networks = [0, 0]\n",
118+
"\n",
119+
"predictions = predictor.predict(inputs, networks=networks)"
120+
]
121+
},
122+
{
123+
"cell_type": "markdown",
124+
"metadata": {},
125+
"source": [
126+
"#### Step 4b: Or give the network a list of inputs to make predictions for and targets to continue learning from\n",
127+
"\n",
128+
"___Note:___ Once retrained, the weights are updated in the object immediately and cannot be undone, unless restored from a previously stored weights file."
129+
]
130+
},
131+
{
132+
"cell_type": "code",
133+
"execution_count": 5,
134+
"metadata": {},
135+
"outputs": [
136+
{
137+
"name": "stdout",
138+
"output_type": "stream",
139+
"text": [
140+
"The target for inputs [1, 71, 0] for network 0 was 106.53 and the prediction was 110.72, error observed was 3.93%. \n",
141+
"The target for inputs [2, 101, 121] for network 0 was 110.72 and the prediction was 115.86, error observed was 4.64%. \n"
142+
]
143+
}
144+
],
145+
"source": [
146+
"inputs = [[1, 71, 0], [2, 101, 121]]\n",
147+
"networks = [0, 0]\n",
148+
"targets = [106.53, 110.72]\n",
149+
"\n",
150+
"predictions = predictor.predict(inputs, networks=networks, targets=targets, recal=True)"
151+
]
152+
},
153+
{
154+
"cell_type": "markdown",
155+
"metadata": {},
156+
"source": [
157+
"The network is now retrained and has updated the weights.\n",
158+
"\n",
159+
"(Observe the difference in the second prediction because the weights were updated after predicting the first value.)"
160+
]
161+
},
162+
{
163+
"cell_type": "markdown",
164+
"metadata": {},
165+
"source": [
166+
"#### Step 4c: Or give the network a CSV file of inputs, targets and networks to make predictions for and to continue learning from\n",
167+
"\n",
168+
"___Note:___ Once retrained, the weights are updated in the object immediately and cannot be undone, unless restored from a previously stored weights file."
169+
]
170+
},
171+
{
172+
"cell_type": "code",
173+
"execution_count": 6,
174+
"metadata": {},
175+
"outputs": [
176+
{
177+
"name": "stdout",
178+
"output_type": "stream",
179+
"text": [
180+
"The target for inputs [1, 71, 0] for network 0 was 106.53 and the prediction was 115.19, error observed was 8.13%. \n",
181+
"The target for inputs [2, 101, 121] for network 0 was 114.81 and the prediction was 114.44, error observed was 0.32%. \n",
182+
"The target for inputs [3, 107, 116] for network 0 was 111.63 and the prediction was 113.52, error observed was 1.70%. \n",
183+
"The target for inputs [4, 24, 81] for network 0 was 111.85 and the prediction was 113.78, error observed was 1.72%. \n",
184+
"The target for inputs [5, -69, 39] for network 0 was 111.87 and the prediction was 113.83, error observed was 1.75%. \n"
185+
]
186+
}
187+
],
188+
"source": [
189+
"inputs, networks, targets = read_input_file(INPUT_FILE)\n",
190+
"\n",
191+
"predictions = predictor.predict(inputs[:5], networks=networks[:5], targets=targets[:5], recal=True)"
192+
]
193+
},
194+
{
195+
"cell_type": "markdown",
196+
"metadata": {},
197+
"source": [
198+
"#### Step 5: Save the predictions to a CSV file\n",
199+
"\n",
200+
"Required parameters:\n",
201+
"- `targets`: As above.\n",
202+
"- `predictions`: As above.\n",
203+
"- `output_file`: Location on disk of the output file, with the name."
204+
]
205+
},
206+
{
207+
"cell_type": "code",
208+
"execution_count": 7,
209+
"metadata": {},
210+
"outputs": [],
211+
"source": [
212+
"write_output_to_file(targets, predictions, os.path.join(DATA_LOCATION, 'predictions_2017.csv'))"
213+
]
214+
},
215+
{
216+
"cell_type": "markdown",
217+
"metadata": {},
218+
"source": [
219+
"#### Step 6: Store the weights for future use\n",
220+
"\n",
221+
"Store the network state for future re-use."
222+
]
223+
},
224+
{
225+
"cell_type": "code",
226+
"execution_count": 8,
227+
"metadata": {},
228+
"outputs": [],
229+
"source": [
230+
"predictor.save(os.path.join(DATA_LOCATION, 'daily_rnn_retrained_17.pickle'))"
231+
]
232+
},
233+
{
234+
"cell_type": "markdown",
235+
"metadata": {},
236+
"source": [
237+
"Further documentation about other functions implemented is available within the source files. The shared variables and parameters are documented in `__init__.py`."
238+
]
239+
},
240+
{
241+
"cell_type": "code",
242+
"execution_count": null,
243+
"metadata": {},
244+
"outputs": [],
245+
"source": []
246+
}
247+
],
248+
"metadata": {
249+
"kernelspec": {
250+
"display_name": "Python 3",
251+
"language": "python",
252+
"name": "python3"
253+
},
254+
"language_info": {
255+
"codemirror_mode": {
256+
"name": "ipython",
257+
"version": 3
258+
},
259+
"file_extension": ".py",
260+
"mimetype": "text/x-python",
261+
"name": "python",
262+
"nbconvert_exporter": "python",
263+
"pygments_lexer": "ipython3",
264+
"version": "3.6.5"
265+
}
266+
},
4267
"nbformat": 4,
5268
"nbformat_minor": 2
6269
}

‎predict_example.ipynb

Lines changed: 39 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -4,7 +4,7 @@
44
"cell_type": "markdown",
55
"metadata": {},
66
"source": [
7-
"### Example for running the RNN\n",
7+
"### Example for predicting using the RNN\n",
88
"\n",
99
"This is an example for running the RNN and get water demand predictions.\n",
1010
"\n",
@@ -15,7 +15,7 @@
1515
"To run this Notebook locally:\n",
1616
"- Jupyter (https://jupyter.org/)\n",
1717
"\n",
18-
"___Note:___ Current model is an online prediction model. The current version of code does __not__ provide an option for training."
18+
"___Note:___ This model is an online prediction model. The current version of code does provides a separate option for training the RNN. Refer to [training_example.ipynb](./training_example.ipynb) for training the RNN."
1919
]
2020
},
2121
{
@@ -56,7 +56,7 @@
5656
"cell_type": "markdown",
5757
"metadata": {},
5858
"source": [
59-
"#### Step 3: To initialize the predictor RNN, create a `recurrent_neural_network` object.\n",
59+
"#### Step 3: To initialize the predictor RNN, create a `rnn_predict` object.\n",
6060
"\n",
6161
"Required parameters:\n",
6262
"- `weights`: Location on disk of the weights file. Typically a Python serialized object (Pickle) file.\n",
@@ -88,7 +88,7 @@
8888
"- `inputs`: A list of `[day_of_the_year, maximum_temperature, precipitation]` lists. These should be real world observed or predicted values.\n",
8989
"\n",
9090
"Optional parameters:\n",
91-
"- `networks`: A list of the same length as `inputs` list with values of the range `[0, 1]`, specifying what network each prediction to run on, typically `num_month - 1`. If not specified, all predictions will be made for the month of January.\n",
91+
"- `networks`: A list of the same length as `inputs` list with values of the range `[0, 11]`, specifying what network each prediction to run on, typically `num_month - 1`. If not specified, all predictions will be made for the month of January.\n",
9292
"- `targets`: A list of the same length as `inputs` of the observed values for the day. The target values are required for updating the weights of the model.\n",
9393
"- `recal`: Boolean value specifying if the network weights should be recalculated. Recommended if using observed demand as target, to improve future predictions. If set to `True`, requires `targets` to be specified. Default value: `False`.\n",
9494
"\n",
@@ -123,7 +123,7 @@
123123
"cell_type": "markdown",
124124
"metadata": {},
125125
"source": [
126-
"#### Step 4b: Or give the network a list of inputs to make predictions for and targets to learn from\n",
126+
"#### Step 4b: Or give the network a list of inputs to make predictions for and targets to continue learning from\n",
127127
"\n",
128128
"___Note:___ Once retrained, the weights are updated in the object immediately and cannot be undone, unless restored from a previously stored weights file."
129129
]
@@ -159,6 +159,38 @@
159159
"(Observe the difference in the second prediction because the weights were updated after predicting the first value.)"
160160
]
161161
},
162+
{
163+
"cell_type": "markdown",
164+
"metadata": {},
165+
"source": [
166+
"#### Step 4c: Or give the network a CSV file of inputs, targets and networks to make predictions for and to continue learning from\n",
167+
"\n",
168+
"___Note:___ Once retrained, the weights are updated in the object immediately and cannot be undone, unless restored from a previously stored weights file."
169+
]
170+
},
171+
{
172+
"cell_type": "code",
173+
"execution_count": 6,
174+
"metadata": {},
175+
"outputs": [
176+
{
177+
"name": "stdout",
178+
"output_type": "stream",
179+
"text": [
180+
"The target for inputs [1, 71, 0] for network 0 was 106.53 and the prediction was 115.19, error observed was 8.13%. \n",
181+
"The target for inputs [2, 101, 121] for network 0 was 114.81 and the prediction was 114.44, error observed was 0.32%. \n",
182+
"The target for inputs [3, 107, 116] for network 0 was 111.63 and the prediction was 113.52, error observed was 1.70%. \n",
183+
"The target for inputs [4, 24, 81] for network 0 was 111.85 and the prediction was 113.78, error observed was 1.72%. \n",
184+
"The target for inputs [5, -69, 39] for network 0 was 111.87 and the prediction was 113.83, error observed was 1.75%. \n"
185+
]
186+
}
187+
],
188+
"source": [
189+
"inputs, networks, targets = read_input_file(INPUT_FILE)\n",
190+
"\n",
191+
"predictions = predictor.predict(inputs[:5], networks=networks[:5], targets=targets[:5], recal=True)"
192+
]
193+
},
162194
{
163195
"cell_type": "markdown",
164196
"metadata": {},
@@ -173,7 +205,7 @@
173205
},
174206
{
175207
"cell_type": "code",
176-
"execution_count": 6,
208+
"execution_count": 7,
177209
"metadata": {},
178210
"outputs": [],
179211
"source": [
@@ -191,7 +223,7 @@
191223
},
192224
{
193225
"cell_type": "code",
194-
"execution_count": 7,
226+
"execution_count": 8,
195227
"metadata": {},
196228
"outputs": [],
197229
"source": [

0 commit comments

Comments
 (0)