<< Click to Display Table of Contents >> Navigation:  Tutorials > Tutorial 10: Structure learning > Tutorial10.py

import pysmile

 

# Tutorial10 loads Credit10k.csv file

# and runs multiple structure learning algorithms

# using the loaded dataset.

# Use the link below to download the Credit10k.csv file:

# https://support.bayesfusion.com/docs/Examples/Learning/Credit10K.csv

 

class Tutorial10:

    def __init__(self):

        print("Starting tutorial10...")

        ds = pysmile.learning.DataSet()

        try:

            ds.read_file("Credit10k.csv")

        except pysmile.SMILEException:

            print("Dataset load failed")

            return

        print(f"Dataset has {ds.get_variable_count()} variables (columns) " 

            + f"and {ds.get_record_count()} records (rows)")

        bayes_search = pysmile.learning.BayesianSearch()

        bayes_search.set_iteration_count(50)

        bayes_search.set_rand_seed(9876543)

        try:

            net1 = bayes_search.learn(ds)

        except pysmile.SMILEException:

            print("Bayesian Search failed")

            return

        print(f"1st Bayesian Search finished, structure score: {bayes_search.get_last_score()}")

        net1.write_file("tutorial10-bs1.xdsl")

 

        bayes_search.set_rand_seed(3456789)

        try:

            net2 = bayes_search.learn(ds)

        except pysmile.SMILEException:

            print("Bayesian Search failed")

            return

        print(f"2nd Bayesian Search finished, structure score: {bayes_search.get_last_score()}")

        net2.write_file("tutorial10-bs2.xdsl")

 

        idx_age = ds.find_variable("Age")

        idx_profession = ds.find_variable("Profession")

        idx_credit_worthiness = ds.find_variable("CreditWorthiness")

 

        if idx_age < 0 or idx_profession < 0 or idx_credit_worthiness < 0:

            print("Can't find dataset variables for background knowledge")

            print("The loaded file may not be Credit10k.csv")

            return

        background_knowledge = pysmile.learning.BkKnowledge()

        background_knowledge.match_data(ds)

        background_knowledge.add_forbidden_arc(idx_age, idx_credit_worthiness)

        background_knowledge.add_forced_arc(idx_age, idx_profession)

 

        bayes_search.set_bk_knowledge(background_knowledge)

        try:

            net3 = bayes_search.learn(ds)

        except pysmile.SMILEException:

            print("Bayesian Search failed")

            return

        print(f"3rd Bayesian Search finished, structure score: {bayes_search.get_last_score()}")

        net3.write_file("tutorial10-bs3.xdsl")

 

        tan = pysmile.learning.TAN()

        tan.set_rand_seed(777999)

        tan.set_class_variable_id("CreditWorthiness")

        try:

            net4 = tan.learn(ds)

        except pysmile.SMILEException:

            print("TAN failed")

            return

        print("Tree-augmented Naive Bayes finished")

        net4.write_file("tutorial10-tan.xdsl")

 

        pc = pysmile.learning.PC()

        try:

            pattern = pc.learn(ds)

        except pysmile.SMILEException:

            print("PC failed")

            return

        net5 = pattern.make_network(ds)

        print("PC finished, proceeding to parameter learning")

        net5.write_file("tutorial10-pc.xdsl")

        em = pysmile.learning.EM()

        try:

            matching = ds.match_network(net5)

        except pysmile.SMILEException:

            print("Can't automatically match network with dataset")

            return

        em.set_uniformize_parameters(False)

        em.set_randomize_parameters(False)

        em.set_eq_sample_size(0)

        try:

            em.learn(ds, net5, matching)

        except pysmile.SMILEException:

            print("EM failed")

            return

        print("EM finished")

        net5.write_file("tutorial10-pc-em.xdsl")

 

        print("Tutorial10 complete.")