Scribd
Upload
0 views

MachineLearningCheatSheet

The document outlines a comprehensive workflow for data preprocessing, visualization, machine learning model selection, evaluation, and optimization using techniques like GridSearch and Pipelines. It includes steps for cleaning data, visualizing results with plots, applying various machine learning classifiers, and performing cross-validation. Additionally, it discusses ensemble methods such as bagging and boosting for improved model performance.

Uploaded by

Bart Mania
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
0 views

MachineLearningCheatSheet

The document outlines a comprehensive workflow for data preprocessing, visualization, machine learning model selection, evaluation, and optimization using techniques like GridSearch and Pipelines. It includes steps for cleaning data, visualizing results with plots, applying various machine learning classifiers, and performing cross-validation. Additionally, it discusses ensemble methods such as bagging and boosting for improved model performance.

Uploaded by

Bart Mania
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
You are on page 1/ 4

1.

Clean data
- verwijder onnodige/overbodige kolommen
- bepaal o.a. shape en unieke waardes van belangrijke kolommen (df.shape,
df.info, df.value_counts)

- overview maken import ydata_profiling


profile = ydata_profiling.ProfileReport(dt)
profile

- label encoding voor class


preprocessing.LabelEncoder()

2. Visualize
- melten en plotten: p = pd.DataFrame(X, columns = labels)
p['class'] = y

# prepare for plotting


p = p.melt(id_vars='class')

plt.figure(figsize=(10,5))
sns.boxplot(data = p, x = 'variable', y='value',
hue='class')

- Facetgrid: g = sns.FacetGrid(p, col='class', row='variable',


height=4)
var = np.unique(p.variable)
g.map(sns.histplot, 'value', kde=True)

3. ML1
- bepaal soort ML (KNeighbors, RandomForestClassifier, DummyClassifier,
LogisticRegression, DecisionTreeClassifier)

- splitten: from sklearn.model_selection import train_test_split


X = df.drop(['class'], axis=1)
y = df['class']
le = preprocessing.LaberEncoder()
y_label = le.fit(y)
y_transform = y_label.transform(y)
X_train, X_test, y_train, y_test = train_test_split(X, y,
test_size=0.x, randomstate=...)

eventueel nogmaals voor train set:


X_train, X_validation, y_train, y_validation =
train_test_split(X_trainval, y_trainval, random_state=1)

- scalen: from sklearn.preprocessing import StandardScaler


scaler = StandardScaler()
y = df['class'].values
data = df.drop('class', axis=1)
labels = data.columns
X = scaler.fit_transform(data)

- fitten voorbeeld:
'classifier' = neighbors.KNeighborsClassifier()
'classifier'.fit(X_train, y_train)

- evalueren/testen 'classifier'.score(X_test, y_test)


from sklearn.metrics import confusion_matrix
yhat = 'classifier'.predict(X_test)
ytrue = y_test
labels = ['class']
cfm = confusion_matrix(y_test, yhat, labels=labels)
cfm

clf_'classifier' = 'classifier'.fit(X_train, y_train)


print('Accuracy of the Decision Tree Classifier on train
set: {:.2f}'.format(clf_'classifier'.score(X_train, y_train)))
print('Accuracy of the Decision Tree Classifier on test
set: {:.2f}'.format(clf_'classifier'.score(X_test, y_test)))

from sklearn.metrics import classification_report,


accuracy_score
print(accuracy_score(y_test, predictions))
report = classification_report(y_test, predictions,
target_names = iris.target_names)
print(report)

- cross validation cross_val_score('classifier', X, y, cv=5)

kfold = model_selection.KFold(n_splits=10)
'classifier'CrossResults =
model_selection.cross_val_score('classifier', X, y_transform, cv=kfold)
'classifier2'CrossResults =
model_selection.cross_val_score('classifier2', X, y_transform, cv=kfold)

sns.boxplot(data=['classifier'CrossResults,
'classifier2'CrossResults])
plt.xticks([0, 1], [''classifier'', ''classifier2''])
plt.xlabel('Model')
plt.ylabel('Accuracy')
plt.title('Cross validated accuracies')
plt.show()

4. GridSearch
- bepalen parameters voor model (e.g. C (regularization parameter) of gamma
(kernel bandwidth))

_____________________________________________________________________________

best_score = 0
for gamma in [0.001, 0.01, 0.1, 1, 10, 100]:
for C in [0.001, 0.01, 0.1, 1, 10, 100]:
# Initialize SVC model for given combination of
parameters
svm = SVC(gamma=gamma, C=C)
# Train on train set
svm.fit(X_train, y_train)
# Evaluate on validation set
score = svm.score(X_validation, y_validation)
# Store the best score
if score > best_score:
best_score = score
best_parameters = {'C': C, 'gamma':
gamma}
_____________________________________________________________________________

svm = SVC(**best_parameters)
# Train on train/validation set
svm.fit(X_trainval, y_trainval)
# Score on test set
test_score = svm.score(X_test, y_test)

_____________________________________________________________________________

- Gridsearch ook ingebouwd

param_grid = {'C': [0.001, 0.01, 0.1, 1, 10, 100],


'gamma': [0.001, 0.01, 0.1, 1, 10, 100]}

grid_search = GridSearchCV(SVC(), param_grid, cv=3)

cv = 2 aangeraden!!!!

Voor beste parameters:


grid_search.best_param_

Voor beste score:


grid_search.best_score_

Voor Cross Validation resultaten:


grid_search.cv_results_

- visualizeren results = pd.DataFrame(grid_search.cv_results_)


# Reshape test scores and plot heatmap
scores = np.array(results.mean_test_score).reshape(7, 6)
ax = sns.heatmap(scores, xticklabels=param_grid['gamma'],
yticklabels=param_grid['C'], annot=True)
ax.set(xlabel='gamma', ylabel='C')

5. Pipeline
- sklearn object dat stappen volgt uit een toegevoegde [LIJST] met (TUPLES)
Iedere tuple bevat gespecificeerde naam en instance van
estimator.
e.g. Pipeline([("scaler", MinMaxScaler()), ("svm", SVC())])

- .fit(X_train, y_train) aanroepen op Pipeline voor trainen model


.score(X_test, y_test) aanroepen voor score

- Pipeline kan in een GridSearch worden gedaan


Wanneer je dit doet, moet je in de Dictionary met parameter
grids (para_grid) aangeven bij welke stap in de Pipeline deze hoort
Voorbeeld:
param_grid = {'svm__C': [0.001, 0.01, 0.1, 1, 10,
100],
'svm__gamma': [0.001, 0.01, 0.1, 1,
10, 100]}
Deze is dus voor de "svm" estimator

GridSearchCV('pipeline', param_grid=param_grid, cv=2)

#. Ensemble, Baggin, Boosting


- ensemble VotingClassifier(['lijst van estimators'])
- bagging BaggingClassifier(estimator='model',
n_estimators='number_of_trees')
RandomForestClassifier(n_estimators='number_of_trees',
max_features='max_features')
ExtraTreesClassifier(n_estimators='number_of_trees',
max_features='max_features')
- boosting AdaBoostClassifier(n_estimators='number_of_trees')
GradientBoostingClassifier(n_estimators='number_of_trees')

XG boost

You might also like