Inside dis second classification lesson, you go explore more ways to classify numeric data. You go also learn about wetin fit happen if you choose one classifier pass the oda one.
We dey assume say you don finish di previous lessons and you get cleaned dataset for your data folder wey dem call cleaned_cuisines.csv for di root of dis 4-lesson folder.
We don load your notebook.ipynb file wit di cleaned dataset and we don divide am into X and y dataframes, ready for di model building process.
Before now, you don learn about di different options wey you get wen you dey classify data using Microsoft cheat sheet. Scikit-learn get similar, but more detailed cheat sheet wey fit help you narrow down your estimators (another word for classifiers):
Tip: visit dis map online and click along di path to read documentation.
Dis map go help wella if you done sabi your data well, because you fit 'walk' along e paths go make decision:
- We get >50 samples
- We want predict category
- We get labeled data
- We get less than 100K samples
- ✨ We fit choose Linear SVC
- If dis no work, since we get numeric data
- We fit try ✨ KNeighbors Classifier
- If dat one no work, try ✨ SVC and ✨ Ensemble Classifiers
- We fit try ✨ KNeighbors Classifier
Dis na better way wey you fit follow.
Follow dis path, we suppose start by import some libraries to use.
-
Import di libraries wey you need:
from sklearn.neighbors import KNeighborsClassifier from sklearn.linear_model import LogisticRegression from sklearn.svm import SVC from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier from sklearn.model_selection import train_test_split, cross_val_score from sklearn.metrics import accuracy_score,precision_score,confusion_matrix,classification_report, precision_recall_curve import numpy as np
-
Split your training and test data:
X_train, X_test, y_train, y_test = train_test_split(cuisines_features_df, cuisines_label_df, test_size=0.3)
Support-Vector clustering (SVC) na pikin from di Support-Vector machines family of ML techniques (you fit learn more about dem below). For dis method, you fit choose 'kernel' to decide how you go cluster di labels. Di 'C' parameter mean 'regularization' wey dey control how parameters go influence di model. Di kernel fit be one of plenty; for here we set am to 'linear' make we use linear SVC. Probability set to 'false' by default; but here we set am to 'true' to get probability estimates. We set random state to '0' to shuffle di data so that we fit get probabilities.
Start by creating array of classifiers. You go dey add to dis array as we dey test.
-
Start with Linear SVC:
C = 10 # Make different classifier dem. classifiers = { 'Linear SVC': SVC(kernel='linear', C=C, probability=True,random_state=0) }
-
Train your model using di Linear SVC and print report:
n_classifiers = len(classifiers) for index, (name, classifier) in enumerate(classifiers.items()): classifier.fit(X_train, np.ravel(y_train)) y_pred = classifier.predict(X_test) accuracy = accuracy_score(y_test, y_pred) print("Accuracy (train) for %s: %0.1f%% " % (name, accuracy * 100)) print(classification_report(y_test,y_pred))
Di result good:
Accuracy (train) for Linear SVC: 78.6% precision recall f1-score support chinese 0.71 0.67 0.69 242 indian 0.88 0.86 0.87 234 japanese 0.79 0.74 0.76 254 korean 0.85 0.81 0.83 242 thai 0.71 0.86 0.78 227 accuracy 0.79 1199 macro avg 0.79 0.79 0.79 1199 weighted avg 0.79 0.79 0.79 1199
K-Neighbors na part of di "neighbors" family of ML methods, wey fit dey used for both supervised and unsupervised learning. For dis method, dem go set how many points before and gather data around dem points so that you fit predict labels wey fit generalize for di data.
Di previous classifier good and e work well for di data, but maybe we fit get better accuracy. Try K-Neighbors classifier.
-
Add one line to your classifier array (put comma after the Linear SVC item):
'KNN classifier': KNeighborsClassifier(C),
Di result just small worse:
Accuracy (train) for KNN classifier: 73.8% precision recall f1-score support chinese 0.64 0.67 0.66 242 indian 0.86 0.78 0.82 234 japanese 0.66 0.83 0.74 254 korean 0.94 0.58 0.72 242 thai 0.71 0.82 0.76 227 accuracy 0.74 1199 macro avg 0.76 0.74 0.74 1199 weighted avg 0.76 0.74 0.74 1199✅ Learn about K-Neighbors
Support-Vector classifiers na part of di Support-Vector Machine family of ML methods wey dem dey use for classification and regression tasks. SVMs "go map training examples go points for space" to make distance between two categories max. Data wey come after go also map inside dis space so that dem fit predict di category.
Make we try small better accuracy with Support Vector Classifier.
-
Add comma after K-Neighbors item then add dis line:
'SVC': SVC(),
Di result good well!
Accuracy (train) for SVC: 83.2% precision recall f1-score support chinese 0.79 0.74 0.76 242 indian 0.88 0.90 0.89 234 japanese 0.87 0.81 0.84 254 korean 0.91 0.82 0.86 242 thai 0.74 0.90 0.81 227 accuracy 0.83 1199 macro avg 0.84 0.83 0.83 1199 weighted avg 0.84 0.83 0.83 1199✅ Learn about Support-Vectors
Make we follow di path reach last, even though di previous test good well. Make we try some 'Ensemble Classifiers, especially Random Forest and AdaBoost:
'RFST': RandomForestClassifier(n_estimators=100),
'ADA': AdaBoostClassifier(n_estimators=100)Di result correct wella, especially for Random Forest:
Accuracy (train) for RFST: 84.5%
precision recall f1-score support
chinese 0.80 0.77 0.78 242
indian 0.89 0.92 0.90 234
japanese 0.86 0.84 0.85 254
korean 0.88 0.83 0.85 242
thai 0.80 0.87 0.83 227
accuracy 0.84 1199
macro avg 0.85 0.85 0.84 1199
weighted avg 0.85 0.84 0.84 1199
Accuracy (train) for ADA: 72.4%
precision recall f1-score support
chinese 0.64 0.49 0.56 242
indian 0.91 0.83 0.87 234
japanese 0.68 0.69 0.69 254
korean 0.73 0.79 0.76 242
thai 0.67 0.83 0.74 227
accuracy 0.72 1199
macro avg 0.73 0.73 0.72 1199
weighted avg 0.73 0.72 0.72 1199
✅ Learn about Ensemble Classifiers
Dis Machine Learning method "go join di predictions from many base estimators" to make di model better. For our example, we use Random Trees and AdaBoost.
-
Random Forest, averaging method, dey build 'forest' of 'decision trees' wey get random nature to avoid overfitting. Di n_estimators parameter na di number of trees.
-
AdaBoost dey fit classifier to dataset then e fit many copies of dat classifier to di same dataset. E dey focus for weights of items wey classifier no classify well and e fit adjust di fit for next classifier to correct am.
Each of these techniques get plenti parameters wey you fit change. Research how each one their default parameters be and reason how changing dem fit affect di model quality.
Plenty big big words dey dis lessons, so take small time review dis list of useful terms!
Disclaimer:
Dis document don translate wit AI translation service Co-op Translator. Even though we dey try make am correct, abeg sabi say automated translation fit get some errors or mistake. Di original document for im own language na di correct source. If na important information, make you use professional human translation. We no go take responsibility for any wrong understanding or mistake wey fit happen because of this translation.