Documentation
Access all resources and documentation or contact us for additional assistance.
Build and run
Most of the time, the only command you need to type is “brainome ”.
For example:
brainome https://download.brainome.ai/data/public/titanic_train.csv
Brainome will automatically (1) clean the data, (2) split it into training and validation sets, (3) extract measurements and select a machine learning model type (DT, NN or RF), (4) build the model based on the measurements, and (5) train the model. At the end, Brainome will present validation measurements that help you understand the quality of the predictions. More on this later.
The output of Brainome is a self-contained predictor that runs with Python3. By default, the predictor’s name is a.py. The predictor itself has a usage help. For example:
python3 a.py
usage: a.py [-h] [-validate] [-cleanfile] [-headerless] [-json] [-trim]
csvfile
a.py: error: the following arguments are required: csvfile
To validate the predictor on the training data set, you can run this command:
python3 a.py -validate titanic_train.csv
Classifier Type: Random Forest
System Type: Binary classifier
Accuracy:
Best-guess accuracy: 61.50%
Model accuracy: 91.12% (729/800 correct)
Improvement over best guess: 29.62% (of possible 38.5%)
Model capacity (MEC): 11 bits
Generalization ratio: 63.72 bits/bit
Confusion Matrix:
Actual | Predicted
—————————-
died | 460 32
survived | 39 269
Accuracy by Class:
target | TP FP TN FN TPR TNR PPV NPV F1 TS
——– | — — — — ——- ——- ——- ——- ——- ——-
died | 460 39 269 32 93.50% 87.34% 92.18% 89.37% 92.84% 86.63%
survived | 269 32 460 39 87.34% 93.50% 89.37% 92.18% 88.34% 79.12%
Exploring the predictor
The predictor is a standalone Python3 program that can be easily inspected. It contains all measurements, including the Brainome command line (“Invocation”) that was used to create the predictor. Here’s the Python predictor that Brainome creates:
more a.py
#!/usr/bin/env python3
#
# This code has been produced by a paid version of Brainome Table Compiler(tm) licensed to: User 1.
# Portions of this code copyright (c) 2019-2021 by Brainome, Inc. All Rights Reserved.
# Brainome grants an exclusive (subject to our continuing rights to use and modify models),
# worldwide, non-sublicensable, and non-transferable limited license to use and modify this
# predictor produced through the input of your data:
# (i) for users accessing the service through a free evaluation account, solely for your
# own non-commercial purposes, including for the purpose of evaluating this service, and
# (ii) for users accessing the service through a paid, commercial use account, for your
# own internal and commercial purposes.
# Please contact support@brainome.ai with any questions.
# Use of predictions results at your own risk.
#
# Output of Brainome Table Compiler v1.002.
# Invocation: brainome https://download.brainome.ai/data/public/titanic_train.csv
# Total compiler execution time: 0:00:13.88. Finished on: May-18-2021 17:55:22.
# This source code requires Python 3.
#
“””
Predictor: a.py
Classifier Type: Random Forest
System Type: Binary classifier
Training / Validation Split: 60% : 40%
Accuracy:
Best-guess accuracy: 61.50%
Training accuracy: 100.00% (479/479 correct)
Validation Accuracy: 77.88% (250/321 correct)
Combined Model Accuracy: 91.12% (729/800 correct)
Model Capacity (MEC): 11 bits
Generalization Ratio: 41.84 bits/bit
Percent of Data Memorized: 4.85%
Resilience to Noise: -1.64 dB
System Meter Runtime Duration: 1s
Training Confusion Matrix:
Actual | Predicted
—— | ———
died | 295 0
survived | 0 184
Validation Confusion Matrix:
Actual | Predicted
—— | ———
died | 165 32
survived | 39 85
Training Accuracy by Class:
Survived | TP FP TN FN TPR TNR PPV NPV F1 TS
——– | —- —- —- —- ——– ——– ——– ——– ——– ——–
died | 295 0 184 0 100.00% 100.00% 100.00% 100.00% 100.00% 100.00%
survived | 184 0 295 0 100.00% 100.00% 100.00% 100.00% 100.00% 100.00%
Validation Accuracy by Class:
Survived | TP FP TN FN TPR TNR PPV NPV F1 TS
——– | —- —- —- —- ——– ——– ——– ——– ——– ——–
died | 165 39 85 32 83.76% 68.55% 80.88% 72.65% 82.29% 69.92%
survived | 85 32 165 39 68.55% 83.76% 72.65% 80.88% 70.54% 54.49%
Attribute Ranking:
Feature | Relative Importance
Sex : 0.4287
Cabin_Class : 0.1785
Cabin_Number : 0.0853
Age : 0.0535
Fare : 0.0465
Sibling_Spouse : 0.0432
Ticket_Number : 0.0407
Port_of_Embarkation : 0.0405
PassengerId : 0.0354
Name : 0.0343
Parent_Children : 0.0133
“””
…
Now that you’ve created and run a predictor, we are ready for some more advanced uses of Brainome.