Depth Of Tree Research Articles

Estimation of soil moisture using decision tree regression

Soil moisture (SM) is a significant factor in the climate system. The accurate determination of SM has high importance in food production to satisfy the increasing demand for food and the chemical processes of soil. This paper applies decision tree regression to estimate SM considering different parameters including air temperature, time, relative humidity, and soil temperature. The presented method holds a mighty advantage to determine SM since the stimulant of the decision tree regression is an algorithm that generates a decision tree from given instances. Besides, usage of decision tree regression provides an opportunity to save time. Numerical results show that the presented method offers a high coefficient of determination value (R2), low mean squared error (MSE), and mean absolute error (MAE). The depth of the decision tree equals to five by providing higher fitness values than other depth levels. The best fitness values in the training stage are 0.00019, 0.007, and 0.842 for MSE, MAE, and R2, respectively. In conclusion of the paper, applied decision tree regression can handle the data of SM estimation in satisfying fitness criterion.

Pressure drops of fresh cemented paste backfills through coupled test loop experiments and machine learning techniques

In this paper, test loop experiments and machine learning techniques were combined to investigate pressure drops of fresh cemented paste backfill (CPB) mixes. The influence of tailings characteristics on CPB pressure drops was studied and Extensive test loop experiments were performed. The complex mapping from tailings characteristics, cement-tailings ratio, inlet velocity and solids content to pressure drop was successfully learned by decision tree regression (DTR) models. In addition, the influence of training set size and maximum tree depth on DTR performance was investigated. The relative importance of predictor variables was discussed and the visualisation for a representative DTR model was provided. Finally, the current research applied two ensemble techniques, namely random forest and gradient boosting regression tree, to increase the predictive performance of the DTR models. The study found that all ensemble techniques outperformed DTR in the pressure drop prediction of CPB.

A Fast-Converging Ensemble Infilling Approach Balancing Global Exploration and Local Exploitation: The Go-Inspired Hybrid Infilling Strategy

AbstractInfilling strategies have been proposed for decades and are widely used in engineering problems. It is still challenging to achieve an effective trade-off between global exploration and local exploitation. In this paper, a novel decision-making infilling strategy named the Go-inspired hybrid infilling (Go-HI) strategy is proposed. The Go-HI strategy combines multiple individual infilling strategies, such as the mean square error (MSE), expected improvement (EI), and probability of improvement (PoI) strategies. The Go-HI strategy consists of two major parts. In the first part, a tree-like structure consisting of several subtrees is built. In the second part, the decision value for each subtree is calculated using a cross-validation (CV)-based criterion. Key factors that significantly influence the performance of the Go-HI strategy, such as the number of component infilling strategies and the tree depth, are explored. Go-HI strategies with different component strategies and tree depths are investigated and also compared with four baseline adaptive sampling strategies through three numerical functions and one engineering case. Results show that the number of component infilling strategies exerts a larger influence on the global and local performance than the tree depth; the Go-HI strategy with two component strategies performs better than the ones with three; the Go-HI strategy always outperforms the three component infilling strategies and the other four benchmark strategies in global performance and robustness and saves much computational cost.

Multiscale habitat mediates pest reduction by birds in an intensive agricultural region

AbstractHabitat augmentation on farms is predicted to conserve biological diversity and support beneficial animals that reduce crop pests. Effectiveness of local habitat enhancement and subsequent pest reduction services can be mediated by the amount of habitat at larger scales. We tested whether the presence and increase of local and landscape scale bird habitat increased avian predator abundance and pest reduction by birds. We surveyed birds and performed a sentinel prey exclosure experiment in walnut orchards in the Sacramento Valley, California, USA—comparing predation probability between orchards with (n = 10) and without (n = 10) woody habitat in uncultivated orchard margins. We digitized seminatural habitat cover in landscapes around orchards to test the effectiveness of avian predators in reducing sentinel prey along a seminatural cover gradient of 0–38%. Experimental prey were diapausing larvae of Cydia pomonella (L.) (Lepidoptera: Tortricidae; codling moth), a significant pest of walnuts, which overwinter in cocoons in orchards, emerge as adults, and produce larvae that feed on the nuts the following spring. Permitting bird access to cocoons increased larval predation from 11% (caged) to 46% (no cage), and predation increased with increasing proportions of seminatural habitat within 500‐m of orchard transects. Predation also increased as the size and bark furrow depth of walnut trees increased, likely because these characteristics were associated with increasing abundance of avian predators with functional traits specific to consuming tree‐dwelling cocoons (e.g., woodpeckers). The presence and increasing complexity of local margin habitat increased the species richness and abundance of avian predators but was not predictive of cocoon predation. Consistent with intermediate landscape‐complexity hypothesis predictions, the effect size of woodpecker abundance on predation was large in simple landscapes (1–20% seminatural cover) and low in complex landscapes (>20% cover). Contrary to predictions, effect size was large in cleared landscapes (<1% cover), suggesting that orchards supported predators in cleared landscapes, with positive effects on pest reduction. We provide evidence that increasing the abundance of avian predators with traits specific for consuming target pests—by retaining old trees and seminatural cover—can increase orchard pest reduction services in an intensive agricultural region.

Machine Learning based Test Case Prioritization in Object Oriented Testing

Software maintenance is one of the most expensive activities in software life cycle. It costs nearly 70% of the total cost of the software. Either to adopt the new requirement or to correct the functionality, software undergoes maintenance. As a consequent of maintenance activities, software undergoes many reforms. Newly added software components may affect the working of existing components and also may introduce faults in existing components. The regression testing tries to reveal the faults that might have been introduced due to these reformations. Running all the prior existing test cases may not be feasible due to constraints like time, cost and resources. Test case prioritization may help in ordered execution of test cases. Running a faulty or fault prone component early in testing process may help in revealing more faults per unit of time. And hence may reduce the testing time. There have been many different criteria for assigning the priority to test cases. But none of the approaches so far have considered the object oriented design metrics for determining the priority of test cases. Object oriented design metrics have been empirically studied for their impact of software maintainability, reliability, testability and quality but usage of these metrics in test case prioritization is still an open area of research. The research reported in this paper evaluates subset of CK metrics. Metrics considered from CK suite include Coupling between objects (CBO), Depth of Inheritance tree (DIT), weighted methods per class (WMC), Number of children (NOC), and Response for a class (RFC). Study also considers four other metrics namely publically inherited methods (PIM), weighted attributes per class (WAC), number of methods inherited (NMI) and number of methods overridden. A model is built based on these metrics for the prediction of software quality and based on the quality measures software modules are classified with the help of Support Vector Machine (SVM) algorithm. The proposed approach is implemented in WEKA tool and analysed on experimental data extracted from open source software. Proposed work would firstly help the tester in identifying the low quality modules and then prioritize the test cases based on quality centric approach. The work also attempts to automate test case prioritization in object oriented testing. The results obtained are encouraging.

From Optimization-Based Machine Learning to Interpretable Security Rules for Operation

Various supervised machine learning approaches have been used in the past to assess the power system security (also known as reliability). This is typically done by training a classifier on a large number of operating points whose postfault status (stable or unstable) has been determined via time-domain simulations. The output of this training process can be expressed as a security rule that is used online to classify an operating point. A critical, and little-studied aspect of these approaches is the interpretability of the rules produced. The lack of interpretability is a well-known issue of some machine learning approaches, especially when dealing with difficult classification problems. In the case of the security assessment of the power system, which is a complex mission-critical task, interpretability is a key requirement for the adoption and deployment by operators of these approaches. In this paper, for the first time, we explore the tradeoff between predictive accuracy and interpretability in the context of power system security assessment. We begin by demonstrating how decision trees (DTs) can be used to learn data-driven security rules and use the tree depth as a measure for interpretability. We leverage disjunctive programming to formulate novel training methods, capable of learning high-quality DTs while still maintaining interpretability. In particular, we propose two new approaches: 1) optimal classification trees is proposed for training DTs of low-depth and 2) greedy optimization-based tree is proposed for training DTs of intermediate depth, where the increased computational burden is managed by exploiting the nested tree structure. We also demonstrate that the ability to generate high-quality interpretable rules can actually translate to impressive benefits in terms of training requirements. Through case studies on the IEEE 68-bus system, we demonstrate that the proposed methods can produce DTs of higher quality compared to the state-of-the-art approach classification and regression tree approach, also if the DT was trained on a significant smaller database, resulting in computational savings of 80%. Given that generating a large training database is a practical bottleneck in these data-driven approaches, this is a significant breakthrough for real-world application.

Risk stratification for low extremity amputation in critical limb ischemia patients who have undergone endovascular revascularization: A survival tree analysis.

Patients with peripheral artery disease (PAD) are a heterogeneous population and differ in risk of mortality and low extremity amputation (LEA), which complicates clinical decision-making. This study aimed to develop a simple risk scale using decision tree methodology to guide physicians in managing critical limb ischemia (CLI) patients who will benefit from endovascular therapy (EVT).A total of 736 patients with CLI, Rutherford classification (RC) stage ≥4, and prior successful EVT were included. Variables significantly associated with LEA by univariate analysis (P < .05) were selected and put into classification tree analysis using the Classification and Regression Tree (CART) model with a dependent variable, amputation, and depth of tree = 3. Four risk groups were generated according to the order of amputation rate. The amputation-free survival (AFS) times between groups were compared using the Kaplan-Meier curve with the log-rank test.Patients were classified as high risk for amputation (G4) (WBC counts ≥10,000/μl, and platelet-lymphocyte ratio (PLR) ≥130.337); intermediate risk group 1 (G3) (WBC < 10,000/μl and RC stage before EVT > 5); intermediate risk group 2 (G2) (WBC count ≥ 10,000/μl, and PLR < 130.337) and low-risk group (G1) (WBC < 10,000/μl, RC before EVT ≤ 5). G2, G3, and G4 risk groups had shorter AFS time (range, 58.7 to 65.5 months) than the G1 risk group (100 months) (P < .05). Risk of LEA was significantly higher in the G4, G3, and G2 groups than in the G1 group (P ≤ .05). The G4 group had the highest risk of amputation (odds ratio = 6.84, P < .001).This simple risk scale model can help healthcare professionals more easily identify and appropriately treat patients with CLI who are at different levels of risk for LEA following endovascular revascularization.

The Organization of Point Cloud Data Based on the Compact Octree Model

While using the octree model to organize a huge amount of scattered point cloud data, it is easy to generate an unbalanced tree structure. Thus, a compact octree model is proposed. It reconstructs the minimum bounding cuboid of each child node to shrink its spatial scope after each division, which effectively decreases the tree depth of scattered point cloud data. Through a series of tests, we verify the feasibility of our proposed spatial index model.

How to Combine Tree-Search Methods in Reinforcement Learning

Finite-horizon lookahead policies are abundantly used in Reinforcement Learning and demonstrate impressive empirical success. Usually, the lookahead policies are implemented with specific planning methods such as Monte Carlo Tree Search (e.g. in AlphaZero (Silver et al. 2017b)). Referring to the planning problem as tree search, a reasonable practice in these implementations is to back up the value only at the leaves while the information obtained at the root is not leveraged other than for updating the policy. Here, we question the potency of this approach. Namely, the latter procedure is non-contractive in general, and its convergence is not guaranteed. Our proposed enhancement is straightforward and simple: use the return from the optimal tree path to back up the values at the descendants of the root. This leads to a γh-contracting procedure, where γ is the discount factor and h is the tree depth. To establish our results, we first introduce a notion called multiple-step greedy consistency. We then provide convergence rates for two algorithmic instantiations of the above enhancement in the presence of noise injected to both the tree search stage and value estimation stage.

Bootstrap Estimated Uncertainty of the Environment Model for Model-Based Reinforcement Learning

Model-based reinforcement learning (RL) methods attempt to learn a dynamics model to simulate the real environment and utilize the model to make better decisions. However, the learned environment simulator often has more or less model error which would disturb making decision and reduce performance. We propose a bootstrapped model-based RL method which bootstraps the modules in each depth of the planning tree. This method can quantify the uncertainty of environment model on different state-action pairs and lead the agent to explore the pairs with higher uncertainty to reduce the potential model errors. Moreover, we sample target values from their bootstrap distribution to connect the uncertainties at current and subsequent time-steps and introduce the prior mechanism to improve the exploration efficiency. Experiment results demonstrate that our method efficiently decreases model error and outperforms TreeQN and other stateof-the-art methods on multiple Atari games.

An Indoor Passive Positioning Method Using CSI Fingerprint Based on Adaboost

Wi-Fi fingerprint-based indoor positioning systems have extensive application prospects. Indoor positioning based on channel state information (CSI) becomes a research hotspot because it can provide fine-grained information. This paper proposes the Adaboost positioning system (ABPS) that uses the phase information in CSI and the ensemble learning (EL) method to train the fingerprint map. In this system, the abnormal phase data is eliminated by density-based clustering and the remainder is linearly transformed to build the fingerprint map. Through continuous iteration with the Adaboost algorithm, the sample weights of the training sets are continuously adjusted to prepare for classification. Finally, it can achieve position coordinates regression by means of confidence level. A series of experiments have been conducted to illustrate the effects of EL parameters, such as the number of iterations, the maximum depth of decision tree, the maximum number of features, and the minimum number of samples required for node subdivision. Meanwhile, the influence of input data sets size is investigated in an open environment and a complex laboratory environment. The experimental results of various classification and regression methods are also discussed, which validates the effectiveness of the proposed method.

Quantitative characterization of rodent feto-placental vasculature morphology in micro-computed tomography images

Background and ObjectiveOptimal development of placental vasculature is critical for fetal growth and health outcomes. Many studies characterizing the vascular structure of the fetal side of the placenta have utilized a range of two-dimensional and three-dimensional (3D) imaging techniques including X-ray micro-computed tomography (micro-CT) following perfusion of the vasculature with a radio-opaque compound. The CT approach has been used to study feto-placental vasculature in rodents and humans. Its inherent advantage is that it reveals the 3D structure in high resolution without destroying the sample. This permits both multiple scanning of the sample and follow-up histological investigations in the same sample. Nevertheless, the applicability of the approach is hampered both by the challenging segmentation of the vasculature and a lack of straightforward methodology to quantitate the feto-placental vascular network. This paper addresses these challenges. MethodsAn end-to-end methodology is presented for automatically segmenting the vasculature; obtaining a Strahler-ordered rooted-tree representation and extracting quantitative features from its nodes, segments and branches (including volume, length, tortuosity and branching angles). The methodology is demonstrated for rat and mouse placentas at the end of gestation (day 22 and day 18, respectively), perfused with Microfil® and imaged using two different micro-CT scanners. ResultsThe 3D visualizations of the resulting vascular trees clearly demonstrate differences between the branching complexity, tree span and tree depth of the mouse and rat placentas. The quantitative characterizations of these trees include not only the fundamental features that have been utilized in other studies of feto-placental vasculature but also several additional features. Boxplots of several of these—tortuosity, number of side branches, number of offspring per branch and branch volume—computed at each Strahler order are presented and interpreted. Differences and similarities between the mouse and rat casts are readily detected. ConclusionThe proposed end-to-end methodology, and the implementation presented using a combination of Amira and Matlab, offers researchers in the field of placental vasculature characterization a straightforward and objective approach for quantifying micro-CT vascular datasets.

Increasing the views and reducing the depth in random forest

Random forest (RF) is one of the most powerful ensemble methods with high performance when dealing with high dimensional data. In this paper, we present a novel approach, inspired by multi-view theory and by human multi-view approach in object recognition, to increase the performance of RF by increasing the number of trees and limiting the number of levels for each tree in RF. The standard RF algorithm constructs the trees to a maximum possible depth until no further classification is needed, or no further feature is available. In the proposed method, trees of RF are bounded to a certain depth to allow increasing the views. That is why this approach is called “Increasing the Views and Reducing the Depth (IVRD)” approach in RF. Each bounded tree is considered as a local view of the problem and the more local view means the better classification/recognition capability. IVRD approach in RF has been tested on a group of synthetic datasets, a standard dataset, and two datasets of Down Syndrome subjects. Several evaluations have been conducted to determine the effects of the tree depth and the number of trees in a forest on the accuracy of IVRD approach. The results showed by bounding the number of levels in each tree and increasing the number of trees in the forest, the accuracy can be improved, especially in very high dimensional problems.

A multi-sensor data fusion enabled ensemble approach for medical data from body sensor networks

Wireless Body Sensor Network (BSNs) are wearable sensors with varying sensing, storage, computation, and transmission capabilities. When data is obtained from multiple devices, multi-sensor fusion is desirable to transform potentially erroneous sensor data into high quality fused data. In this work, a data fusion enabled Ensemble approach is proposed to work with medical data obtained from BSNs in a fog computing environment. Daily activity data is obtained from a collection of sensors which is fused together to generate high quality activity data. The fused data is later input to an Ensemble classifier for early heart disease prediction. The ensembles are hosted in a Fog computing environment and the prediction computations are performed in a decentralised manners. The results from the individual nodes in the fog computing environment are then combined to produce a unified output. For the classification purpose, a novel kernel random forest ensemble is used that produces significantly better quality results than random forest. An extensive experimental study supports the applicability of the solution and the obtained results are promising, as we obtain 98% accuracy when the tree depth is equal to 15, number of estimators is 40, and 8 features are considered for the prediction task.

Thermal Comfort-Based Personalized Models with Non-Intrusive Sensing Technique in Office Buildings

Heating, ventilation and air-conditioning (HVAC) systems play a key role in shaping the built environment. However, centralized HVAC systems cannot guarantee the provision of a comfortable thermal environment for everyone. Therefore, a personalized HVAC system that aims to adapt thermal preferences has drawn much more attention. Meanwhile, occupant-related factors like skin temperature have not had standardized measurement methods. Therefore, this paper proposes to use infrared thermography to develop individual thermal models to predict thermal sensations using three different feature sets with the random forest (RF) and support vector machine (SVM). The results have shown the correlation coefficients between clothing surface temperature and thermal sensation are 11% and 3% higher than those between skin temperature and thermal sensation of two subjects, respectively. With cross-validation, SVM with a linear kernel and penalty number of 1, as well as RF with 50 trees and the maximum tree depth of 3 were selected as the model configurations. As a result, the model trained with the feature set, consisting of indoor air temperature, relative humidity, skin temperature and clothing surface temperature, and with linear kernel SVM has achieved 100% recall score on test data of female subjects and 95% recall score on that of male subjects.

Attribute Selection Based on Constraint Gain and Depth Optimal for a Decision Tree.

Uncertainty evaluation based on statistical probabilistic information entropy is a commonly used mechanism for a heuristic method construction of decision tree learning. The entropy kernel potentially links its deviation and decision tree classification performance. This paper presents a decision tree learning algorithm based on constrained gain and depth induction optimization. Firstly, the calculation and analysis of single- and multi-value event uncertainty distributions of information entropy is followed by an enhanced property of single-value event entropy kernel and multi-value event entropy peaks as well as a reciprocal relationship between peak location and the number of possible events. Secondly, this study proposed an estimated method for information entropy whose entropy kernel is replaced with a peak-shift sine function to establish a decision tree learning (CGDT) algorithm on the basis of constraint gain. Finally, by combining branch convergence and fan-out indices under an inductive depth of a decision tree, we built a constraint gained and depth inductive improved decision tree (CGDIDT) learning algorithm. Results show the benefits of the CGDT and CGDIDT algorithms.

Towards Intelligent Mining for Backfill: A genetic programming-based method for strength forecasting of cemented paste backfill

As cemented paste backfill (CPB) plays an increasingly important role in minerals engineering, forecasting its mechanical properties becomes a necessity for efficient CPB design. Machine learning (ML) techniques have previously demonstrated remarkable successes in such task by providing black-box predictions. To express the non-linear relationship in an explicit and precise way, we employed genetic programming (GP) for the uniaxial compressive strength (UCS) prediction of CPB. The influence of sampling method, training set size and maximum tree depth on the GP performance was investigated. A detailed analysis was conducted on a representative GP model and the relative variable importance was investigated using the relative variable frequency, partial dependence plots and relative importance scores. The statistical parameters show that a satisfactory performance was obtained by the GP modelling (R2 > 0.80 on the testing set). Results of this study indicate that cement-tailings ratio, solids content and curing time were the most three important variables for the UCS prediction. The predictive performance of GP modelling was comparable to well-recognised ML techniques, and the trained GP model can be generalised to entirely new tailings with satisfactory performance. This study indicates that the GP-based method is capable of providing explicit and precise forecasting of UCS, which can serve as a reliable tool for quick, inexpensive and effective assessment of UCS in the absence of adequate experimental data.

Depth Limitation and Splitting Criteria Optimization on Random Forest for Efficient Human Activity Classification

Random Forest (RF) is known as one of the best classifiers in many fields. They are parallelizable, fast to train and to predict, robust to outlier, handle unbalanced data, have low bias, and moderate variance. Apart from these advantages, there are still opportunities to increase RF efficiency. The absence of recommendations regarding the number of trees involved in RF ensembles could make the number of trees very large. This can increase the computational complexity of RF. Recommendations for not pruning the decision tree further aggravates the condition. This research attempts to build an efficient RF ensemble while maintaining its accuracy, especially in problem activity. Data collection is performed using an accelerometer sensor on a smartphone device. The data used in this research are collected from five peoples who perform 11 different activities. Each activity is carried out five times to enrich the data. This study uses two steps to improve the efficiency of the classification of the activity: 1) Optimal splitting criteria for activity classification, 2) Measured pruning to limit the tree depth in RF ensemble. The first method in this study can be applied to determine the splitting criteria that are most suitable for the classification problem of activities using Random Forest. In this case, the decision model built using the Gini Index can produce the highest accuracy. The second method proposed in this research successfully builds less complex pruned-tree without reducing its classification accuracy. The research results showed that the method applied to the Random Forest in this study was able to produce a decision model that was simple but yet accurate to classify activity.

Effects of Tree Root Density on Soil Total Porosity and Non-Capillary Porosity Using a Ground-Penetrating Tree Radar Unit in Shanghai, China

To explore the relationship between arbor root density and corresponding physical soil properties (total porosity, non-capillary porosity, specific weight, bulk density, moisture content, and permeability), noninvasive methods such as Tree Radar Unit (TRU) were utilized to detect the root systems of 10 species of tree that are commonly used in Shanghai green spaces with more than 70% frequency of occurrence. Using TRU and soil investigation, root density and soil porous features were determined for different tree species, depth slices, and distances. The relationships among root density, soil porosity, and non-capillary porosity were identified using variance analysis and regression analysis. The results indicated that root density decreased with increasing distance from the trunk and soil depth. Soil porosity and non-capillary porosity had significantly positive and linear correlations with root density. Compared with lawns, trees such as Zelkova serrata, Koelreuteria paniculata, Cinnamomum camphora, and Metasequoia glyptostroboides had the greatest effects on soil improvement through soil porosity and non-capillary porosity. Due to the spatial distribution of root systems, trees could be divided into three types based on their influence at various soil depths.

Microbial Community Structure in a Malaysian Tropical Peat Swamp Forest: The Influence of Tree Species and Depth.

Tropical peat swamp forests sequester globally significant stores of carbon in deep layers of waterlogged, anoxic, acidic and nutrient-depleted peat. The roles of microbes in supporting these forests through the formation of peat, carbon sequestration and nutrient cycling are virtually unknown. This study investigated physicochemical peat properties and microbial diversity between three dominant tree species: Shorea uliginosa (Dipterocarpaceae), Koompassia malaccensis (legumes associated with nitrogen-fixing bacteria), Eleiodoxa conferta (palm) and depths (surface, 45 and 90 cm) using microbial 16S rRNA gene amplicon sequencing. Water pH, oxygen, nitrogen, phosphorus, total phenolic contents and C/N ratio differed significantly between depths, but not tree species. Depth also strongly influenced microbial diversity and composition, while both depth and tree species exhibited significant impact on the archaeal communities. Microbial diversity was highest at the surface, where fresh leaf litter accumulates, and nutrient supply is guaranteed. Nitrogen was the core parameter correlating to microbial communities, but the interactive effects from various environmental variables displayed significant correlation to relative abundance of major microbial groups. Proteobacteria was the dominant phylum and the most abundant genus, Rhodoplanes, might be involved in nitrogen fixation. The most abundant methanogens and methanotrophs affiliated, respectively, to families Methanomassiliicoccaceae and Methylocystaceae. Our results demonstrated diverse microbial communities and provide valuable insights on microbial ecology in these extreme ecosystems.