Analysis of Soft Decision Trees for Passive-Expert Reinforcement Learning

This paper explores the use of soft decision trees [1] in basic reinforcement applications to examine the efficacy of using passive-expert like networks for optimal Q-Value learning on Artificial Neural Networks (ANN). The soft decision tree networks were built using the PyTorch machine learning and the OpenAi’s Gym environment frameworks. The conducted research study aimed at assessing the performance of soft decision tree networks on Cartpole as provided in the OpenAi Gym software package. The baseline performance metric that the soft decision tree networks were compared against was a simple Deep Neural Network using several linear layers with ReLU and Softmax activation functions for the input and output layers, respectively. All networks were trained using the Backpropagation algorithm provided generically by PyTorch’sAutograd module.

Ordinal Decision Trees

In many decision making tasks,the features and decision are ordinal.Several ordinal classification learning algorithms have been developed in recent years,it is shown that these algorithms are sensitive to noisy samples and do not work in real-world applications.In this work,we propose a new measure of feature quality, called rank mutual information.Then,we design an ordinal decision tree(REOT) construction technique based on rank mutual information.The theoretic and experimental analysis shows that the proposed algorithm is effective.

Analysis of NIR spectroscopic data using decision trees and their ensembles

Decision trees and their ensembles became quite popular for data analysis during the past decade.One of the main reasons for that is current boom in big data,where traditional statistical methods(such as,e.g.,multiple linear regression)are not very efficient.However,in chemometrics these methods are still not very widespread,first of all because of several limitations related to the ratio between number of variables and observations.This paper presents several examples on how decision trees and their ensembles can be used in analysis of NIR spectroscopic data both for regression and classification.We will try to consider all important aspects including optimization and validation of models,evaluation of results,treating missing data and selection of most important variables.The performance and outcome of the decision tree-based methods are compared with more traditional approach based on partial least squares.

Selecting decision trees for power system security assessment

Power systems transport an increasing amount of electricity,and in the future,involve more distributed renewables and dynamic interactions of the equipment.The system response to disturbances must be secure and predictable to avoid power blackouts.The system response can be simulated in the time domain.However,this dynamic security assessment(DSA)is not computationally tractable in real-time.Particularly promising is to train decision trees(DTs)from machine learning as interpretable classifiers to predict whether the systemwide responses to disturbances are secure.In most research,selecting the best DT model focuses on predictive accuracy.However,it is insufficient to focus solely on predictive accuracy.Missed alarms and false alarms have drastically different costs,and as security assessment is a critical task,interpretability is crucial for operators.In this work,the multiple objectives of interpretability,varying costs,and accuracies are considered for DT model selection.We propose a rigorous workflow to select the best classifier.In addition,we present two graphical approaches for visual inspection to illustrate the selection sensitivity to probability and impacts of disturbances.We propose cost curves to inspect selection combining all three objectives for the first time.Case studies on the IEEE 68 bus system and the French system show that the proposed approach allows for better DT-selections,with an 80%increase in interpretability,5%reduction in expected operating cost,while making almost zero accuracy compromises.The proposed approach scales well with larger systems and can be used for models beyond DTs.Hence,this work provides insights into criteria for model selection in a promising application for methods from artificial intelligence(AI).

A Preliminary Study of Interpreting CNNs Using Soft Decision Trees

Decision trees can be used to enhance the interpretability of neural networks.In this work,we compare the classification and interpretability performance of the normal decision tree and a type of soft decision tree when they are used to interpret the decision paths of CNN networks.With the help of feature visualization and human-labeled features,we demonstrate that the soft decision trees identify more consistent features while maintaining much higher classification performance than the normal decision tree.

Clustering feature decision trees for semi-supervised classification from high-speed data streams

Most stream data classification algorithms apply the supervised learning strategy which requires massive labeled data.Such approaches are impractical since labeled data are usually hard to obtain in reality.In this paper,we build a clustering feature decision tree model,CFDT,from data streams having both unlabeled and a small number of labeled examples.CFDT applies a micro-clustering algorithm that scans the data only once to provide the statistical summaries of the data for incremental decision tree induction.Micro-clusters also serve as classifiers in tree leaves to improve classification accuracy and reinforce the any-time property.Our experiments on synthetic and real-world datasets show that CFDT is highly scalable for data streams while gener-ating high classification accuracy with high speed.

Self-Tuning Parameters for Decision Tree Algorithm Based on Big Data Analytics

Big data is usually unstructured, and many applications require theanalysis in real-time. Decision tree (DT) algorithm is widely used to analyzebig data. Selecting the optimal depth of DT is time-consuming process as itrequires many iterations. In this paper, we have designed a modified versionof a (DT). The tree aims to achieve optimal depth by self-tuning runningparameters and improving the accuracy. The efficiency of the modified (DT)was verified using two datasets (airport and fire datasets). The airport datasethas 500000 instances and the fire dataset has 600000 instances. A comparisonhas been made between the modified (DT) and standard (DT) with resultsshowing that the modified performs better. This comparison was conductedon multi-node on Apache Spark tool using Amazon web services. Resultingin accuracy with an increase of 6.85% for the first dataset and 8.85% for theairport dataset. In conclusion, the modified DT showed better accuracy inhandling different-sized datasets compared to standard DT algorithm.

A Data-Driven Oil Production Prediction Method Based on the Gradient Boosting Decision Tree Regression

Accurate prediction ofmonthly oil and gas production is essential for oil enterprises tomake reasonable production plans,avoid blind investment and realize sustainable development.Traditional oil well production trend prediction methods are based on years of oil field production experience and expertise,and the application conditions are very demanding.With the rapid development of artificial intelligence technology,big data analysis methods are gradually applied in various sub-fields of the oil and gas reservoir development.Based on the data-driven artificial intelligence algorithmGradient BoostingDecision Tree(GBDT),this paper predicts the initial single-layer production by considering geological data,fluid PVT data and well data.The results show that the GBDT algorithm prediction model has great accuracy,significantly improving efficiency and strong universal applicability.The GBDTmethod trained in this paper can predict production,which is helpful for well site optimization,perforation layer optimization and engineering parameter optimization and has guiding significance for oilfield development.

Recognition of Hybrid PQ Disturbances Based on Multi-Resolution S-Transform and Decision Tree

Aiming at the problems of multiple types of power quality composite disturbances,strong feature correlation and high recognition error rate,a method of power quality composite disturbances identification based on multiresolution S-transform and decision tree was proposed.Firstly,according to IEEE standard,the signal models of seven single power quality disturbances and 17 combined power quality disturbances are given,and the disturbance waveform samples are generated in batches.Then,in order to improve the recognition accuracy,the adjustment factor is introduced to obtain the controllable time-frequency resolution through multi-resolution S-transform time-frequency domain analysis.On this basis,five disturbance time-frequency domain features are extracted,which quantitatively reflect the characteristics of the analyzed power quality disturbance signal,which is less than the traditional method based on S-transform.Finally,three classifiers such as K-nearest neighbor,support vector machine and decision tree algorithm are used to effectively complete the identification of power quality composite disturbances.Simulation results showthat the classification accuracy of decision tree algorithmis higher than that of K-nearest neighbor and support vector machine.Finally,the proposed method is compared with other commonly used recognition algorithms.Experimental results show that the proposedmethod is effective in terms of detection accuracy,especially for combined PQ interference.

Research on the Intelligent Distribution System of College Dormitory Based on the Decision Tree Classification Algorithm

The trend toward designing an intelligent distribution system based on students’individual differences and individual needs has taken precedence in view of the traditional dormitory distribution system,which neglects the students’personality traits,causes dormitory disputes,and affects the students’quality of life and academic quality.This paper collects freshmen's data according to college students’personal preferences,conducts a classification comparison,uses the decision tree classification algorithm based on the information gain principle as the core algorithm of dormitory allocation,determines the description rules of students’personal preferences and decision tree classification preferences,completes the conceptual design of the database of entity relations and data dictionaries,meets students’personality classification requirements for the dormitory,and lays the foundation for the intelligent dormitory allocation system.

A Dimensional Reduction Approach Based on Essential Constraints in Linear Programming

This paper presents a new dimension reduction strategy for medium and large-scale linear programming problems. The proposed method uses a subset of the original constraints and combines two algorithms: the weighted average and the cosine simplex algorithm. The first approach identifies binding constraints by using the weighted average of each constraint, whereas the second algorithm is based on the cosine similarity between the vector of the objective function and the constraints. These two approaches are complementary, and when used together, they locate the essential subset of initial constraints required for solving medium and large-scale linear programming problems. After reducing the dimension of the linear programming problem using the subset of the essential constraints, the solution method can be chosen from any suitable method for linear programming. The proposed approach was applied to a set of well-known benchmarks as well as more than 2000 random medium and large-scale linear programming problems. The results are promising, indicating that the new approach contributes to the reduction of both the size of the problems and the total number of iterations required. A tree-based classification model also confirmed the need for combining the two approaches. A detailed numerical example, the general numerical results, and the statistical analysis for the decision tree procedure are presented.

A Data and Knowledge Collaboration Strategy for Decision-Making on the Amount of Aluminum Fluoride Addition Based on Augmented Fuzzy Cognitive Maps

In the aluminum reduction process, aluminum uoride (AlF3) is added to lower the liquidus temperature of the electrolyte and increase the electrolytic ef ciency. Making the decision on the amount of AlF3 addi- tion (referred to in this work as MDAAA) is a complex and knowledge-based task that must take into con- sideration a variety of interrelated functions;in practice, this decision-making step is performed manually. Due to technician subjectivity and the complexity of the aluminum reduction cell, it is dif cult to guarantee the accuracy of MDAAA based on knowledge-driven or data-driven methods alone. Existing strategies for MDAAA have dif culty covering these complex causalities. In this work, a data and knowl- edge collaboration strategy for MDAAA based on augmented fuzzy cognitive maps (FCMs) is proposed. In the proposed strategy, the fuzzy rules are extracted by extended fuzzy k-means (EFKM) and fuzzy deci- sion trees, which are used to amend the initial structure provided by experts. The state transition algo- rithm (STA) is introduced to detect weight matrices that lead the FCMs to desired steady states. This study then experimentally compares the proposed strategy with some existing research. The results of the comparison show that the speed of FCMs convergence into a stable region based on the STA using the proposed strategy is faster than when using the differential Hebbian learning (DHL), particle swarm optimization (PSO), or genetic algorithm (GA) strategies. In addition, the accuracy of MDAAA based on the proposed method is better than those based on other methods. Accordingly, this paper provides a feasible and effective strategy for MDAAA.

Retrieval of Antarctic sea ice freeboard and thickness from HY-2B satellite altimeter data

Antarctic sea ice is an important part of the Earth’s atmospheric system,and satellite remote sensing is an important technology for observing Antarctic sea ice.Whether Chinese Haiyang-2B(HY-2B)satellite altimeter data could be used to estimate sea ice freeboard and provide alternative Antarctic sea ice thickness information with a high precision and long time series,as other radar altimetry satellites can,needs further investigation.This paper proposed an algorithm to discriminate leads and then retrieve sea ice freeboard and thickness from HY-2B radar altimeter data.We first collected the Moderate-resolution Imaging Spectroradiometer ice surface temperature(IST)product from the National Aeronautics and Space Administration to extract leads from the Antarctic waters and verified their accuracy through Sentinel-1 Synthetic Aperture Radar images.Second,a surface classification decision tree was generated for HY-2B satellite altimeter measurements of the Antarctic waters to extract leads and calculate local sea surface heights.We then estimated the Antarctic sea ice freeboard and thickness based on local sea surface heights and the static equilibrium equation.Finally,the retrieved HY-2B Antarctic sea ice thickness was compared with the CryoSat-2 sea ice thickness and the Antarctic Sea Ice Processes and Climate(ASPeCt)ship-based observed sea ice thickness.The results indicate that our classification decision tree constructed for HY-2B satellite altimeter measurements was reasonable,and the root mean square error of the obtained sea ice thickness compared to the ship measurements was 0.62 m.The proposed sea ice thickness algorithm for the HY-2B radar satellite fills a gap in this application domain for the HY-series satellites and can be a complement to existing Antarctic sea ice thickness products;this algorithm could provide long-time-series and large-scale sea ice thickness data that contribute to research on global climate change.

Cloud-Based Diabetes Decision Support System Using Machine Learning Fusion

Diabetes mellitus,generally known as diabetes,is one of the most common diseases worldwide.It is a metabolic disease characterized by insulin deciency,or glucose(blood sugar)levels that exceed 200 mg/dL(11.1 ml/L)for prolonged periods,and may lead to death if left uncontrolled by medication or insulin injections.Diabetes is categorized into two main types—type 1 and type 2—both of which feature glucose levels above“normal,”dened as 140 mg/dL.Diabetes is triggered by malfunction of the pancreas,which releases insulin,a natural hormone responsible for controlling glucose levels in blood cells.Diagnosis and comprehensive analysis of this potentially fatal disease necessitate application of techniques with minimal rates of error.The primary purpose of this research study is to assess the potential role of machine learning in predicting a person’s risk of developing diabetes.Historically,research has supported the use of various machine algorithms,such as naïve Bayes,decision trees,and articial neural networks,for early diagnosis of diabetes.However,to achieve maximum accuracy and minimal error in diagnostic predictions,there remains an immense need for further research and innovation to improve the machine-learning tools and techniques available to healthcare professionals.Therefore,in this paper,we propose a novel cloud-based machine-learning fusion technique involving synthesis of three machine algorithms and use of fuzzy systems for collective generation of highly accurate nal decisions regarding early diagnosis of diabetes.

An Ethical Approach to Decision Design

This paper describes an approach to ethical decision design. It discusses the use of several mathematical tools applicable to the world of economics and business. It then demonstrates how the tools analogously extend themselves to making ethical decisions as supported by personal values. A review of the mathematics, including the design of a decision tree, the concept of mathematical expectation, and mathematical modeling are included.

Landslide susceptibility zonation method based on C5.0 decision tree and K-means cluster algorithms to improve the efficiency of risk management

Machine learning algorithms are an important measure with which to perform landslide susceptibility assessments, but most studies use GIS-based classification methods to conduct susceptibility zonation.This study presents a machine learning approach based on the C5.0 decision tree(DT) model and the K-means cluster algorithm to produce a regional landslide susceptibility map. Yanchang County, a typical landslide-prone area located in northwestern China, was taken as the area of interest to introduce the proposed application procedure. A landslide inventory containing 82 landslides was prepared and subsequently randomly partitioned into two subsets: training data(70% landslide pixels) and validation data(30% landslide pixels). Fourteen landslide influencing factors were considered in the input dataset and were used to calculate the landslide occurrence probability based on the C5.0 decision tree model.Susceptibility zonation was implemented according to the cut-off values calculated by the K-means cluster algorithm. The validation results of the model performance analysis showed that the AUC(area under the receiver operating characteristic(ROC) curve) of the proposed model was the highest, reaching 0.88,compared with traditional models(support vector machine(SVM) = 0.85, Bayesian network(BN) = 0.81,frequency ratio(FR) = 0.75, weight of evidence(WOE) = 0.76). The landslide frequency ratio and frequency density of the high susceptibility zones were 6.76/km^(2) and 0.88/km^(2), respectively, which were much higher than those of the low susceptibility zones. The top 20% interval of landslide occurrence probability contained 89% of the historical landslides but only accounted for 10.3% of the total area.Our results indicate that the distribution of high susceptibility zones was more focused without containing more " stable" pixels. Therefore, the obtained susceptibility map is suitable for application to landslide risk management practices.

Canadian children’s and youth’s adherence to the 24-h movement guidelines during the COVID-19 pandemic: A decision tree analysis

Purpose:The purpose of this study was to use decision tree modeling to generate profiles of children and youth who were more and less likely to meet the Canadian 24-h movement guidelines during the coronavirus disease-2019(COVID-19)outbreak.Methods:Data for this study were from a nationally representative sample of 1472 Canadian parents(Meanage=45.12,SD=7.55)of children(5
11 years old)or youth(12
17 years old).Data were collected in April 2020 via an online survey.Survey items assessed demographic,behavioral,social,micro-environmental,and macro-environmental characteristics.Four decision trees of adherence and non-adherence to all movement recommendations combined and each individual movement recommendation(physical activity(PA),screen time,and sleep)were generated.Results:Results revealed specific combinations of adherence and non-adherence characteristics.Characteristics associated with adherence to the recommendation(s)included high parental perceived capability to restrict screen time,annual household income of
CAD 100,000,increases in children’s and youth’s outdoor PA/sport since the COVID-19 outbreak began,being a boy,having parents younger than 43 years old,and small increases in children’s and youth’s sleep duration since the COVID-19 outbreak began.Characteristics associated with non-adherence to the recommendation(s)included low parental perceived capability to restrict screen time,youth aged 12
17 years,decreases in children’s and youth’s outdoor PA/sport since the COVID-19 outbreak began,primary residences located in all provinces except Quebec,low parental perceived capability to support children’s and youth’s sleep and PA,and annual household income ofCAD 99,999.Conclusion:Our results show that specific characteristics interact to contribute to(non)adherence to the movement behavior recommendations.Results highlight the importance of targeting parents’perceived capability for the promotion of children’s and youth’s movement behaviors during challenging times of the COVID-19 pandemic,paying particular attention to enhancing parental perceived capability to restrict screen time.

Risk factor analysis and clinical decision tree model construction for diabetic retinopathy in Western China

BACKGROUND Diabetic retinopathy(DR)is the driving force of blindness in patients with type 2 diabetes mellitus(T2DM).DR has a high prevalence and lacks effective therapeutic strategies,underscoring the need for early prevention and treatment.Yunnan province,located in the southwest plateau of China,has a high prevalence of DR and an underdeveloped economy.AIM To build a clinical prediction model that will enable early prevention and treatment of DR.METHODS In this cross-sectional study,1654 Han population with T2DM were divided into groups without(n=826)and with DR(n=828)based on fundus photography.The DR group was further subdivided into non-proliferative DR(n=403)and proliferative DR(n=425)groups.A univariate analysis and logistic regression analysis were conducted and a clinical decision tree model was constructed.RESULTS Diabetes duration≥10 years,female sex,standing-or supine systolic blood pressure(SBP)≥140 mmHg,and cholesterol≥6.22 mmol/L were risk factors for DR in logistic regression analysis(odds ratio=2.118,1.520,1.417,1.881,and 1.591,respectively).A greater severity of chronic kidney disease(CKD)or hemoglobin A 1c increased the risk of DR in patients with T2DM.In the decision tree model,diabetes duration was the primary risk factor affecting the occurrence of DR in patients with T2DM,followed by CKD stage,supine SBP,standing SBP,and body mass index(BMI).DR classification outcomes were obtained by evaluating standing SBP or BMI according to the CKD stage for diabetes duration<10 years and by evaluating CKD stage according to the supine SBP for diabetes duration≥10 years.CONCLUSION Based on the simple and intuitive decision tree model constructed in this study,DR classification outcomes were easily obtained by evaluating diabetes duration,CKD stage,supine or standing SBP,and BMI.

A New Method for Constructing Decision Tree Based on Rough Sets Theory

Decision trees induction algorithms have been used for classification in a wide range of application domains. In the process of constructing a tree, the criteria of selecting test attributes will influence the classification accuracy of the tree.In this paper,the degree of dependency of decision attribute to condition attribute,based on rough set theory,is used as a heuristic for selecting the attribute that will best separate the samples into individual classes.The result of an example shows that compared with the entropy-based approach,our approach is a better way to select nodes for constructing decision trees.

Research on Scholarship Evaluation System based on Decision Tree Algorithm

Under the modern education system of China, the annual scholarship evaluation is a vital thing for many of the collegestudents. This paper adopts the classification algorithm of decision tree C4.5 based on the bettering of ID3 algorithm and constructa data set of the scholarship evaluation system through the analysis of the related attributes in scholarship evaluation information.And also having found some factors that plays a significant role in the growing up of the college students through analysis and re-search of moral education, intellectural education and culture&PE.