run.py

Go to the documentation of this file.

#!/usr/bin/env python

'''
Ensemble Anomaly Detector Framework

Author:
        Vedanth Narayanan
File:
        Example of training script for ROS module
Date:
        14 May, 2018

'''


import sys
from learn_lib.interface import Interface
# from interface import Interface
import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
from mpl_toolkits.mplot3d.art3d import Poly3DCollection
from sklearn import svm
import yaml
import rospy
from sklearn.decomposition import PCA
import os

def graph_3d(test, y_pred_test):

        print test.shape
        print y_pred_test.shape

        te0, te1, te2 = test[:, 0], test[:, 1], test[:, 2]
        y_pred_test = y_pred_test[0]

        fig = plt.figure()
        ax = fig.gca(projection='3d')

        ax.scatter(te0, te1, te2, c=y_pred_test, \
        cmap='bwr', \
        s=10)

        plt.show()


def reduce_graph_3d(data, preds):

        pca = PCA(n_components=3)

        pca.fit(data)
        data = pca.transform(data)

        graph_3d(data, preds)



if __name__ == "__main__":

        unsupervised_models = []
        supervised_models = []


        print os.getcwd()
        print sys.argv

        input_file = sys.argv[1] #"learning_script.yaml"
        interface = Interface(input_file)
        # interface.print_all()


        with open(input_file, 'r') as fil:
                local = yaml.load(fil)

                fit_file = local["Files"]["fit_file"]
                unsupervised_train_file = local["Files"]["unsupervised_train"]
                supervised_train_file = local["Files"]["supervised_train"]
                testing_file = local["Files"]["testing"]

        clf_ocsvm17 = svm.OneClassSVM(nu=0.05, kernel="rbf", gamma=1000)
        unsupervised_models.append(('ocsvm17', clf_ocsvm17))
        clf_ocsvm18 = svm.OneClassSVM(nu=0.06, kernel="rbf", gamma=1100)
        unsupervised_models.append(('ocsvm18', clf_ocsvm18))


        clf_rbfsvm1 = svm.SVC(kernel='rbf', gamma=150, C=100000)
        supervised_models.append(('clf_rbfsvm1', clf_rbfsvm1))
        # clf_rbfsvmbest_3 = svm.SVC(kernel='rbf', gamma=100, C=100000)
        # supervised_models.append(('rbfsvmbest_3', clf_rbfsvmbest_3))

        interface.genmodel_train(unsupervised_models, supervised_models)

        # TODO: Allow for getting predictions for a particular model
        #               This will ideally be ignored, because learned will be
        #                       in the monitoring phase.
        # new_item = "data/rbfsvm_train_20.csv"
        u_preds, s_preds = interface.get_testing_predictions()

        # print testing_file
        # print testing_file[0]["name"]
        x, y = interface.retrieve_data(testing_file[0]["name"])

        print 'x shape: ', x.shape
        print 'y shape: ', np.array(y).shape

        # print rospy.get_param('unsupervised_model')
        # print rospy.get_param('supervised_model')

        graph_3d(x, s_preds[0])

        # reduce_graph_3d(x, u_preds[0])