Классификации дерева в машинном обучении и гиперпараметрах

Аннотация

Machine learning algorithms play a crucial role in extracting valuable insights from data, enabling businesses and researchers to make informed decisions. One such algorithm is the decision tree, which is widely used for classification tasks. Decision tree classification utilizes a tree-like model of decisions and their potential consequences, making it an intuitive and powerful tool for solving complex problems. In this article, a model that determines which drug is suitable for a 71 patient with a certain disease is created using the Decision tree algorithm. This problem is multi-class classification (multiclass classification) help score consolidation. Alternatively, how function, domain, and hyperparameters simplify decision tree models are explored. Key words: model, dataset, test set, training set, hyperparameters, classification, prediction, decision tree, multiclass classification. The purpose of decision tree classification is to divide a dataset into homogeneous groups based on input features, leading to the assignment of categorical labels to new, unseen instances. This algorithm mimics human decision-making processes by forming a tree structure where each internal node represents a decision based on a feature, and each leaf node represents a class label. Decision trees are versatile and find application in various domains. They are widely used in finance for credit scoring, fraud detection, and risk assessment. In healthcare, decision trees aid in disease diagnosis and treatment prediction. Additionally, decision trees are valuable in customer segmentation, sentiment analysis, and recommendation systems. The decision tree algorithm involves a series of steps to construct an optimal tree structure. It begins with selecting the best feature from the dataset that effectively divides the instances into distinct classes. This process is repeated recursively for each subset of instances until the tree is fully grown. The algorithm uses measures such as information gain, gain ratio, or Gini index to evaluate the feature's effectiveness at splitting the data. Hyperparameters are settings that control the behavior of the machine learning algorithm. In the context of decision trees, hyperparameters influence the tree's structure and complexity. Tuning these hyperparameters can simplify the decision tree model and improve its performance. One common hyperparameter is the maximum depth of the tree, which limits the number of decision nodes and reduces complexity. Lowering the maximum depth helps avoid overfitting. Another crucial hyperparameter is the minimum number of samples required to split an internal node. Increasing this value prevents the creation of small branches that may lead to overfitting. Additionally, decision tree models can be regularized using hyperparameters like minimum impurity decrease or maximum number of leaf nodes. These parameters control the growth of the tree by setting thresholds for stopping the splitting process. Decision Tree Algoritmi Medical data is collected for research. This data (DataSet) is about patients suffering from the same disease. During the course of treatment, one of the 5 different drugs was given to the patient. The goal is to create a model using the Decision tree algorithm that determines which drug may be suitable for a future patient with the same disease. This problem is solved using multiclass classification. The graphic below is an example of how a Decision Tree works. 72 How to build a Decision Tree model • Select a column of the dataset • We consider column importance in data partitioning • We split the data by the best column • We repeat the above steps. Entropy determines which division leads to a better result. https://journal.jbnuu. uz/index.php/ijcstr/article/view/579 The necessary libraries and modules for the Decision Tree algorithm are called. The required dataset is called to build the model. https://raw.githubusercontent.com/JamshidSalimov/Ai-Fayls/master/drug200.csv 73 Textual data is converted to digital form. The predictor and original values are extracted from the dataset and equated to the X and y variables Dataset is split into train_set and test_set using train_test_split() module (60% train_set, 40% test_set) 74 The Decision Tree module is called and trained based on the Dataset. Predictive values are found Based on the predicted values, the model is evaluated Model performance is evaluated using the classification_report() module based on actual values and predicted values It is also possible to calculate these values separately It is also possible to evaluate in matrix form using the confusion_matrix() module. 75 Cross-validation of the model may give better results 76 Based on the values, the Decision Tree algorithm graph was drawn. Hyperparameters`min_impurity_decrease' - defines how "clean" the result will be. The default value is 0 Using the max_dpth parameter, it is possible to control tree branches, that is, tree branch lay