maXbox starter67 machine learning V

////////////////////////////////////////////////////////////////////////////
Machine Learning V
____________________________________________________________________________
maXbox Starter 67 - Data Science with Max
There are two kinds of data scientists:
1) Those who can extrapolate from incomplete data.
This tutor makes a second step with our classifiers in Scikit-Learn of the
preceding tutorial 66 with predicting test sets and histograms.
Lets start with a prediction of one sample:
svm = LinearSVC(random_state=100, C=2) #C for regularization a scale
svm.fit(X,y)
print('predict single sample ',svm.predict([[7,8,9,8]]))
>>> array([1.])
As you can see the sample is unseen data and was not in the dataset before:
A B C D
[[ 1. 2. 3. 4. 0.]
[ 3. 4. 5. 6. 0.]
[ 5. 6. 7. 8. 1.]
[ 7. 8. 9. 10. 1.]
[10. 8. 6. 4. 0.]
[ 9. 7. 5. 3. 1.]]
Next topic is the data split. Normally the data we use is split into training
data and test data. The training set contains a known output and the model
learns on this data alone in order to be generalized to other data later on. We
have a test dataset (or subset) in order to test our model’s prediction on this
subset. We’ll do this by Scikit-Learn library and specifically:
from sklearn.model_selection import train_test_split
But our dataset is very small so we have to split it by 0.5. A test_size=0.2
inside the function indicates a percentage of the data that should be held over
for testing. It’s usually around 80/20 or 70/30.
X_train,X_test,y_train,y_test=train_test_split(X,y,test_size=0.5,random_state=0)
X_train [[7. 8. 9.10.]
[ 1. 2. 3. 4.]
[10. 8. 6. 4.]]
X_test [[9. 7. 5. 3.]
[ 5. 6. 7. 8.]
[ 3. 4. 5. 6.]]
Now we fit the model on the training data:
svm.fit(X_train,y_train)
y_pred = svm.predict(X_test)
print('class test report:
n',classification_report(y_test,y_pred,target_names=targets))
1/7

As you can see, we are fitting the model on the training data and trying to
predict the test data.
class test report:
precision recall f1-score support
class 0 0.50 1.00 0.67 1
class 1 1.00 0.50 0.67 2
avg / total 0.83 0.67 0.67 3
There are two possible predicted classes: 1 as "yes" and 0 as "no". If we were
predicting for example the presence of a disease, "yes" would mean they have the
disease, and "no" would mean they don't have the disease.
The class test Score is 0.66 and the train score is 1.0; its usual that the
train score is higher than the test score to avoid over-fitting.
What about the confusion matrix:
print('Class test Score:', svm.score(X_test, y_test))
print(confusion_matrix(y_test, y_pred))
[[1 0]
[1 1]]
There's a false negative in the 0-Class so the precision is only 0.5!
Best fit would be:
[[1 0]
[0 2]]
Now we plot the model (500 is just the marker bubble size):
plt.scatter(y_test, y_pred, 500)
plt.xlabel('True Values')
plt.ylabel('Predictions')
plt.show()
2/7

Now we want to see how to build a histogram in Python from the ground up.
df= pd.DataFrame.from_records(arr2, columns=features+['Class'])
df.iloc[0:,0:4].plot.hist(density=False)
plt.show()
We want to see how many times our features X[0:4] occur
A B C D
[[ 1. 2. 3. 4.]
[ 3. 4. 5. 6.]
[ 5. 6. 7. 8.]
[ 7. 8. 9. 10.]
[10. 8. 6. 4.]
[ 9. 7. 5. 3.]]
Lets take the feature D the 4 is count 2 times, the 3,6,8,10 counts one time.
Mathematically, a histogram is a mapping of bins (intervals or numbers) to
frequencies. More technically, it can be used to approximate a probability
density function (PDF) of the underlying variable that we see later on.
Moving on from a frequency table above (density=False counts at y-axis), a true
histogram first <bins> the range of values and then counts the number of values
that fall into each bin or interval. Our bins are so small that we can count
each number as one bin. A plot of a histogram uses its bin edges on the x-axis
and the corresponding frequencies on the y-axis.
Now a kernel density estimation (KDE) is a way to estimate the probability
density function (PDF) of the random variable that “underlies” our sample. KDE
is a means of data smoothing and studying density.
print(y,'n',X,'n')
[0. 0. 1. 1. 0. 1.]
[[ 1. 2. 3. 4.]
[ 3. 4. 5. 6.]
[ 5. 6. 7. 8.]
[ 7. 8. 9. 10.]
[10. 8. 6. 4.]
[ 9. 7. 5. 3.]]
df.iloc[0:,0:4].plot.kde()
3/7

Sticking with the Pandas library, you can create and overlay density plots using
plot.kde(), which is available for both [Series] and [DataFrame] objects.
df.iloc[0:,0:4].plot.kde()
This is also possible for our binary targets to see a probabilistic distribution
of the target class values (labels in supervised learning): [0. 0. 1. 1. 0. 1.]
Consider at last a sample of floats drawn from the Laplace and Normal
distribution together. This distribution graph has fatter tails than a normal
distribution and has two descriptive parameters (location and scale):
>>> d = np.random.laplace(loc=15, scale=3, size=500)
>>> d = np.random.normal(loc=15, scale=3, size=500)
4/7

svm = LinearSVC(random_state=100)
y_pred = svm.fit(X,y).predict(X) # fit and predict in one line
print('linear svm score: ',accuracy_score(y, y_pred))
"Classification with Cluster & PCA Example"
Now we will use linear SVC to partition our graph into clusters and split the
data into a training set and a test set for further predictions.
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)
# Run classifier, using a model that is too regularized (C too low) to see
# the impact on the results
classifier = svm.SVC(kernel='linear', C=0.01)
y_pred = classifier.fit(X_train, y_train).predict(X_test)
By setting up a dense mesh of points in the grid and classifying all of them, we
can render the regions of each cluster as distinct colors:
def plotPredictions(clf):
xx, yy = np.meshgrid(np.arange(0, 250000, 10),
np.arange(10, 70, 0.5))
Z = clf.predict(np.c_[xx.ravel(), yy.ravel()])
plt.figure(figsize=(8, 6))
Z = Z.reshape(xx.shape)
plt.contourf(xx, yy, Z, cmap=plt.cm.Paired, alpha=0.8)
plt.scatter(X[:,0], X[:,1], c=y.astype(np.float))
plt.show()
A simple CNN architecture was trained on MNIST dataset using TensorFlow with 1e-
3 learning rate and cross-entropy loss using four different optimizers: SGD,
Nesterov Momentum, RMSProp and Adam.
5/7

We compared different optimizers used in training neural networks and gained
intuition for how they work. We found that SGD with Nesterov Momentum and Adam
produce the best results when training a simple CNN on MNIST data in TensorFlow.
https://sourceforge.net/projects/maxbox/files/Docu/EKON_22_machinelearning_slide
s_scripts.zip/download
Last note concerning PCA and Data Reduction or Factor Analysis:
As PCA simply transforms the input data, it can be applied both to
classification and regression problems. In this section, we will use a
classification task to discuss the method.
The script can be found at:
http://www.softwareschule.ch/examples/811_mXpcatest_dmath_datascience.pas
..examples811_mXpcatest_dmath_datascience.pas
from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis as QA
clf = QA()
y_pred = clf.fit(X,y).predict(X)
print('n QuadDiscriminantAnalysis score2: ',accuracy_score(y, y_pred))
print(confusion_matrix(y, y_pred))
It may be seen that:
• High correlations exist between the original variables, which are
therefore not independent
• The first principal factor is negatively correlated with the second and
fourth variables, and positively correlated with the third variable
• The second principal factor is positively correlated with the first variable
• The table of principal factors show that the highest scores are usually
associated with the first two principal factors, in agreement with the
previous results
Const
N = 6; { Number of observations }
Nvar = 4; { Number of variables }
Of course, its not always this and that simple. Often, we don't know what number
of dimensions is advisable in upfront. In such a case, we leave n_components or
Nvar parameter unspecified when initializing PCA to let it calculate the full
transformation. After fitting the data, explained_variance_ratio_ contains an
6/7

array of ratios in decreasing order: The first value is the ratio of the basis
vector describing the direction of the highest variance, the second value is the
ratio of the direction of the second highest variance, and so on.
print('pearson correlation, coeff:, p-value:')
for i in range(3):
print (pearsonr(X[:,i],X[:,i+1]))
corr = np.corrcoef(X, rowvar=0) # correlation matrix
w, v = np.linalg.eig(corr) # eigen values & eigen vectors
print('eigenvalues & eigenvector:')
print(w) print(v)
>>> eigenvalues & eigenvector:
[ 2.66227922e+00 1.33772078e+00 -4.33219131e-18 6.51846049e-17]
[[-0.46348627 0.56569851 -0.11586592 0.19557765]
[-0.5799298 0.27966438 0.53986421 0.24189689]
[-0.5724941 -0.30865195 -0.71438049 -0.75459654]
[-0.34801208 -0.71169306 0.42986304 0.57777101]]
C:maXboxmX46210DataScienceconfusionlist
You can detect high-multi-collinearity by inspecting the eigen values of
correlation matrix. A very low eigen value shows that the data are collinear,
and the corresponding eigen vector shows which variables are collinear. If there
is no collinearity in the data, you would expect that none of the eigen values
are close to zero.
Being a linear method, PCA has, of course, its limitations when we are faced
with strange data that has non-linear relationships. We wont go into much more
details here, but its sufficient to say that there are extensions of PCA.
The script can be found:
http://www.softwareschule.ch/examples/classifier_compare2confusion2.py.txt
from sklearn.metrics import accuracy_score
from sklearn.metrics import confusion_matrix
from sklearn.metrics import classification_report
from sklearn.model_selection import train_test_split
Ref:
http://www.softwareschule.ch/box.htm
https://scikit-learn.org/stable/modules/
https://realpython.com/python-histograms/
Doc:
https://maxbox4.wordpress.com
7/7

maXbox starter67 machine learning V

More Related Content

What's hot

Similar to maXbox starter67 machine learning V

More from Max Kleiner

Recently uploaded

maXbox starter67 machine learning V