learning.py

#! /usr/bin/python2
# used for DATAFrames and DataFrames can hold different types data of multidimensional arrays.
import pandas as pd
# Numpy provides robust data structures for efficient computation of multi-dimensional arrays & matrices.
import numpy as np
import pickle
import sklearn.ensemble as ske
from sklearn import model_selection, tree, linear_model
from sklearn.feature_selection import SelectFromModel
import joblib
from sklearn.naive_bayes import GaussianNB
from sklearn.metrics import confusion_matrix
from sklearn import tree

data = pd.read_csv('data.csv', sep='|')  # generate df as data
# now droping some coloumns as axis 1(mean coloumn) and will show the values in the rows
X = data.drop(['Name', 'md5', 'legitimate'], axis=1).values
y = data['legitimate'].values  # values of legitimate data

print('Researching important feature based on %i total features\n' %
      X.shape[1])  # shape() is use in pandas to give number of row/column

# Feature selection using Trees Classifier
fsel = ske.ExtraTreesClassifier().fit(X, y)
model = SelectFromModel(fsel, prefit=True)
X_new = model.transform(X)  # now features are only 9 :)
nb_features = X_new.shape[1]  # will save value 13 as shape is (138047, 13) :}

# now converting in training and testing data in 20% range hahhahaha ! as total x is 138047 and testing is 138047*0.2=27610 :)
X_train, X_test, y_train, y_test = model_selection.train_test_split(
    X_new, y, test_size=0.2)
features = []

print('%i features identified as important:' %
      nb_features)  # as mentioned above


# important features sorted
indices = np.argsort(fsel.feature_importances_)[::-1][:nb_features]
for f in range(nb_features):
    print("%d. feature %s (%f)" % (
        f + 1, data.columns[2+indices[f]], fsel.feature_importances_[indices[f]]))

# mean adding to the empty 'features' array the 'important features'
# [::-1] mean start with last towards first
for f in sorted(np.argsort(fsel.feature_importances_)[::-1][:nb_features]):
    features.append(data.columns[2+f])

print(features)

# Algorithm comparison
algorithms = {
    # The max_depth parameter denotes maximum depth of the tree.

    # In case, of random forest, these ensemble classifiers are
    #    the randomly created decision trees. Each decision tree is a single classifier and the target prediction is based on
    #     the majority voting method.
    "RandomForest": ske.RandomForestClassifier(n_estimators=50),
    # n_estimators ==The number of trees in the forest.

    "GradientBoosting": ske.GradientBoostingClassifier(n_estimators=50),
    "AdaBoost": ske.AdaBoostClassifier(n_estimators=100),
    # Ada mean Adaptive
    # Both are boosting algorithms which means that they convert a set of weak learners into a single strong learner. They
    #    both initialize a strong learner (usually a decision tree) and iteratively create a weak learner that is added to the
    #        strong learner. They differ on how they create the weak learners during the iterative process.

    "GNB": GaussianNB(),
    # Bayes theorem is based on conditional probability. The conditional probability helps us calculating the probability
    #        that something will happen
    "DecisionTree": tree.DecisionTreeClassifier(max_depth=10)
}
results = {}
print("\nNow testing algorithms")
for algo in algorithms:
    clf = algorithms[algo]
    clf.fit(X_train, y_train)  # fit may be called as 'trained'
    score = clf.score(X_test, y_test)
    print("%s : %f %%" % (algo, score*100))
    results[algo] = score

tree.plot_tree(clf)

winner = max(results, key=results.get)
print('\nWinner algorithm is %s with a %f %% success' %
      (winner, results[winner]*100))


# Save the algorithm and the feature list for later predictions
print('Saving algorithm and feature list in classifier directory...')
# Persist an arbitrary Python object into one file.
joblib.dump(algorithms[winner], 'classifier/classifier.pkl')
open('classifier/features.pkl', 'wb').write(pickle.dumps(features))
# joblib works especially well with NumPy arrays which are used by sklearn so depending on the classifier type you use you might
# have performance and size benefits using joblib.Otherwise pickle does work correctly so saving a trained classifier and loading
# it again will produce the same results no matter which of the serialization libraries you use
print('Saved')

# Identify false and true positive rates
clf = algorithms[winner]
res = clf.predict(X_test)
mt = confusion_matrix(y_test, res)
print(mt)
# A confusion matrix, also known as an error matrix,[4] is a specific table layout that allows visualization
# of the performance of an algorithm, typically a supervised learning.
print("False positive rate : %f %%" % ((mt[0][1] / float(sum(mt[0])))*100))
print('False negative rate : %f %%' % ((mt[1][0] / float(sum(mt[1]))*100)))