MGI: A New Heuristic for classifying continuous attributes in decision trees

While analyzing the data, it is crucial to choose the model that best matches the circumstance. Many experts in the field of classification and regression have proposed ensemble strategies for tabular data, as well as various approaches to classification and regression problems. In this paper, Gini Index is applied on raw geographical dataset to convert continuous data into discrete dataset. Decision tree algorithm is implemented on resultant discrete dataset, Information Gain is calculated for every attribute and the attribute with highest information gain is the splitting node, applied recursively. Decision tree algorithm implemented predicts the rainfall in Kashmir province with the accuracy of 81.5%. MDL pruning is applied on the resultant decision tree in order to reduce the size & complexity of the Decision tree. Pruning removes segments of the tree that contribute little towards classification; the accuracy is marginally reduced to 81.1%. Furthermore, after the implementation of Decision tree a boosting algorithm: gradient boosting has been implemented on the same set of data using decision tree as a base estimator. It was observed that the overall accuracy of the decision tree got increased to 87.5% after the implementation of gradient boosting model. Thus, the obtained results predict that gradient boosted-DT outperforms all other approaches with the highest accuracy measure and high susceptibility rate in rainfall prediction.

ABSTRACTThis study introduces a new multi‐criteria decision‐making (MCDM) framework to evaluate trauma injury detection models in intensive care units (ICUs). This research addresses the challenges associated with diverse machine learning (ML) models, inconsistencies, conflicting priorities, and the importance of metrics. The developed methodology consists of three phases: dataset identification and pre‐processing, hybrid model development, and an evaluation/benchmarking framework. Through meticulous pre‐processing, the dataset is tailored to focus on adult trauma patients. Forty hybrid models were developed by combining eight ML algorithms with four filter‐based feature‐selection methods and principal component analysis (PCA) as a dimensionality reduction method, and these models were evaluated using seven metrics. The weight coefficients for these metrics are determined using the 2‐tuple Linguistic Fermatean Fuzzy‐Weighted Zero‐Inconsistency (2TLF‐FWZIC) method. The Vlsekriterijumska Optimizcija I Kompromisno Resenje (VIKOR) approach is applied to rank the developed models. According to 2TLF‐FWZIC, classification accuracy (CA) and precision obtained the highest importance weights of 0.2439 and 0.1805, respectively, while F1, training time, and test time obtained the lowest weights of 0.1055, 0.0886, and 0.1111, respectively. The benchmarking results revealed the following top‐performing models: the Gini index with logistic regression (GI‐LR), the Gini index with a decision tree (GI_DT), and the information gain with a decision tree (IG_DT), with VIKOR Q score values of 0.016435, 0.023804, and 0.042077, respectively. The proposed MCDM framework is assessed and examined using systematic ranking, sensitivity analysis, validation of the best‐selected model using two unseen trauma datasets, and mode explainability using the SHapley Additive exPlanations (SHAP) method. We benchmarked the proposed methodology against three other benchmark studies and achieved a score of 100% across six key areas. The proposed methodology provides several insights into the empirical synthesis of this study. It contributes to advancing medical informatics by enhancing the understanding and selection of trauma injury detection models for ICUs.

Classification is an important task in the field of machine learning. Decision tree algorithms are a popular choice for handling classification tasks due to their high accuracy, simple algorithmic process, and good interpretability. Traditional decision tree algorithms, such as ID3, C4.5, and CART, differ primarily in their criteria for splitting trees. Shannon entropy, Gini index, and mean squared error are all examples of measures that can be used as splitting criteria. However, their performance varies on different datasets, making it difficult to determine the optimal splitting criterion. As a result, the algorithms lack flexibility. In this paper, we introduce the concept of generalized entropy from information theory, which unifies many splitting criteria under one free parameter, as the split criterion for decision trees. We propose a new decision tree algorithm called RSE (RS-Entropy decision tree). Additionally, we improve upon a two-term information measure method by incorporating penalty terms and coefficients into the split criterion, leading to a new decision tree algorithm called RSEIM (RS-Entropy Information Method). In theory, the improved algorithms RSE and RSEIM are more flexible due to the presence of multiple free parameters. In experiments conducted on several datasets, using genetic algorithms to optimize the parameters, our proposed RSE and RSEIM methods significantly outperform traditional decision tree methods in terms of classification accuracy without increasing the complexity of the resulting trees.

Decision tree-based state clustering has emerged in as the most popular approach for clustering the states of context dependent hidden Markov model based speech recognizers. The application of sets of phones, mainly phonetically motivated, that limit the possible clusters, results in a reasonably good modeling of unseen phones while it still enables to model specific phones very precisely whenever this is necessary and enough training data is available. Formal concept analysis, a young mathematical discipline, provides means for the treatment of sets and sets of sets that are well suited for further improving tree-based state clustering. The possible refinements are outlined and evaluated in this paper. The major merit is the proposal of procedures for the adaptation of the number of sets used for clustering to the amount of available training data, and of a method that generates suitable sets automatically without the incorporation of additional knowledge.

The favourable growth performance of SSA over the last 20 years (figure 15.1) has been accompanied by a modest decline in poverty, from 59 to 48 per cent over 1993-2010, i.e., much less than that recorded in South Asia (Ferreira, 2014). This aggregate trend, however, conceals substantial cross-country variations. The key question, then, is: How can such differences in poverty reduction rates be explained? The standard approach (Bourguignon, 2003) shows that a percentage change in poverty incidence can be decomposed into the percentage changes in the growth rate of GDP per capita and in the Gini coefficient, plus a small residual. In this regard, it must be noted that in SSA, the average GDP growth per capita oscillated in a narrow range, i.e., between 1.7 per cent in non-resource-rich countries and 2.6 per cent in resource-rich ones. The reason that poverty declined at different rates is, therefore, to be found in the divergence of inequality trends experienced by the SSA countries. Indeed, this chapter and Chapter 2 argue that over 1991-2011, income inequality rose in several countries but declined in a similar number.

Naive Bayes Tree uses decision tree as the general structure and deploys naive Bayesian classifiers at leaves. The intuition is that naive Bayesian classifiers work better than decision trees when the sample data set is small. Therefore, after several attribute splits when constructing a decision tree, it is better to use naive Bayesian classifiers at the leaves than to continue splitting the attributes. In this paper, we propose a learning algorithm to improve the conditional probability estimation in the diagram of Naive Bayes Tree. The motivation for this work is that, for cost-sensitive learning where costs are associated with conditional probabilities, the score function is optimized when the estimates of conditional probabilities are accurate. The additional benefit is that both the classification accuracy and Area Under the Curve (AUC) could be improved. On a large suite of benchmark sample sets, our experiments show that the CLL tree outperforms the state-of-art learning algorithms, such as Naive Bayes Tree and naive Bayes significantly in yielding accurate conditional probability estimation and improving classification accuracy and AUC.

To eliminate the impact on the original data, the visual design sample set is subjected to mathematical transformation in this paper. The Gini coefficient is calculated based on an attribute, the decision tree is trained, and a set of functions is taken to obtain the final output by averaging all decision trees to obtain an unbiased data set estimate and help construct the decision tree. The data analysis resulted in a significant increase in the overall indicators, with the imagery concept visual elements increasing from 60% to 100%, and the overall indicators were above 50%. More visual elements with traditional Chinese cultural meanings should be created in visual design practice and made to become a new carrier for the inheritance and regeneration of non-heritage traditional culture in visual design development.

In response to the low accuracy and recall of current English translation text error recognition methods, this paper proposes a research on English translation text error recognition based on an improved decision tree algorithm. Firstly, use mutual information to calculate the degree of relevance of entries and annotate the part of speech of English translation texts. Then, by using an encoder and decoder to construct a neural network structure, the neural network is applied to the process of feature extraction in machine English translation, and the softmax function is used for normalization. Finally, a fuzzy decision tree is used to segment each information feature, and combined with the Gini index, error features are classified to achieve English translation text error recognition. Through experiments, it has been proven that the accuracy of the translation text error recognition method proposed in this article remains above 94%, the recall rate remains above 86%, the recognition accuracy is high, and the judgment is reliable. The improved decision tree algorithm exhibits consistent performance across varying data volumes, achieving an accuracy exceeding 93% when handling 15,000 to 50,000 sentences, surpassing comparable state-of-the-art algorithms. As such, the enhanced approach for identifying errors in English translation texts effectively enhances translation quality and demonstrates promising potential for practical applications.

Data dimension reduction (DDR) is all about mapping data from high dimensions to low dimensions, various techniques of DDR are being used for image dimension reduction like Random Projections, Principal Component Analysis (PCA), the Variance approach, LSA-Transform, the Combined and Direct approaches, and the New Random Approach. Auto-encoders (AE) are used to learn end-to-end mapping. In this paper, we demonstrate that pre-processing not only speeds up the algorithms but also improves accuracy in both supervised and unsupervised learning. In pre-processing of DDR, first PCA based DDR is used for supervised learning, then we explore AE based DDR for unsupervised learning. In PCA based DDR, we first compare supervised learning algorithms accuracy and time before and after applying PCA. Similarly, in AE based DDR, we compare unsupervised learning algorithm accuracy and time before and after AE representation learning. Supervised learning algorithms including support-vector machines (SVM), Decision Tree with GINI index, Decision Tree with entropy and Stochastic Gradient Descent classifier (SGDC) and unsupervised learning algorithm including K-means clustering, are used for classification purpose. We used two datasets MNIST and FashionMNIST Our experiment shows that there is massive improvement in accuracy and time reduction after pre-processing in both supervised and unsupervised learning.

Leaf rolling is a common response to drought among members of the grass family. A detailed understanding of the structural changes that occur when rice leaves roll in response to drought may assist in determining the physiological mechanisms underlying this feature and determining its potential utility in plant breeding. The anatomical characteristics of rice cultivars native to northeastern Thailand in response to drought stress were evaluated in this study. Polyethylene glycol (PEG) treatments were applied to seedlings of 28 native rice cultivars and two rice breeding cultivars for 10 d. When cultivated in drought-prone environments, the native rice cultivars had a leaf rolling index between 4 and 7, which was classified as moderate to high leaf rolling. Rice leaf anatomy exhibited profound changes in response to drought, with smaller bulliform cells (36.8%), thicker epidermis (10.98%), smaller vascular bundles (7.41%), and smaller bundle sheath cells (11.95%). Decision trees were used to explore the association between the degree of leaf rolling and anatomical traits. According to the decision tree models generated on the basis of the gain ratio, information gain, and Gini index, the epidermis, leaf thickness, and bulliform cells were the major factors of the root nodes, and the maximum accuracy of the models was 70.90%. Overall, the models indicated that rice leaves with a thin epidermis, large bulliform cells, thin leaves, and small vascular bundles are more likely to display high leaf rolling adaptations under drought stress conditions.

The focus of this paper is on diagnosing the diabetes using different supervised machine learning classifiers such as Neural Networks, SVM, KNN, Naive Bayes technique and Decision trees using holdout validation. The diabetic dataset classification is one of the research problems of machine learning research community. The pima Indian diabetes dataset which is available at [1] UCI machine learning repository has been used in all the experiments mentioned in this paper and MATLAB 2014 has been used to perform the experiments. Here we have mainly focus on the performance evaluation methods like accuracy, error rate, sensitivity, specificity, confusion matrix and AUC. Keywords: Diabetes, Diagnosis, Classification, Accuracy, Machine learning.

As customers tendencies learn rapidly about online transactions due to the outbreak of the Covid-19 pandemic, fraudulent transactions that mimic legitimate transactions will continue to increase unabated. Over the years, fraudulent transactions have been committed but the supervised and unsupervised machine learning classifiers and fraud detection techniques could not stop the fraud from occurring at the onset. However, fraudulent cases are detected after post-occurrence thereby recognizing these techniques as non-smart detection models because fraud could only be detected using past data sets. Various fraud detection or classifier techniques using machine learning techniques such as decision trees, random forests, k-means, and neural networks have been studied. Also, the possibility of overfitting due to the branch growth associated with the decision trees is a classifier performance problem. The application of the probability axiomatic metric (PAM) to determine the accuracy of the models is a classification evaluation problem because of the relatively moderate error rate. To overcome the above shortfalls, we proposed a threshold Mahalanobis distance (TMD) classifier and adopted BETH as a performance evaluation metric. The result revealed that TMD is comparable to neural network (NN), k-means, isolation forest, XGBoost, and random forest based on the credit card data set. The analysis showed that TMD and NN have higher efficiency compared to other classifiers. This study demonstrated that the TMD classifier is unique and could be applied to extract legitimate and illegitimate transactions from customer transaction data sets. The study concludes that TMD is robust and comparable to NN and the classification errors associated with BETH are minimal compared to PAM for both TMD and NN.

A scalable decision tree system and its application in pattern recognition and intrusion detection

In this paper we are comparing two methods for classification of Old Slavic Cyrillic characters. Traditional character recognition programs cannot be applied to the Old Slavic Church manuscripts due to the specific characteristics of these characters. The first classification method is based on a decision tree and the second one uses a fuzzy classifier. Both methods employ the same set of features extracted from the characters bitmaps. The prototypes are obtained by applying the logical operators on the samples of digitalized characters from original manuscripts. As features for defining a particular character we use number and position of spots in the outer segments, presence and position of beams and columns in horizontal and vertical segments respectively, compactness and symmetry. The fuzzy classifier creates a prototype consisting of fuzzy rules by means of fuzzy aggregation of character features. The classifier based on a decision tree is realized by a set of rules. During the creation of the classifier, several splitting measures are tested. We have created an application that implements the proposed classifiers and have experimentally tested their efficiency. Experimental results show that both classifiers correctly recognize about 50% of the characters. For 10% of the samples both classifiers make the same error and for 11% of characters the predicted character is incorrect and different.

Comparative Study of Twoing and Entropy Criterion for Decision Tree Classification of Dispersed Data