Locally Weighted Learning Research Articles

Autoencoder-based nonlinear Bayesian locally weighted regression for soft sensor development

The framework of locally weighted learning (LWL) has established itself as a popular tool for developing nonlinear soft sensors in process industries. For LWL-based soft sensors, the key factor for achieving high performance is to construct accurate localized models. To this end, in this paper a nonlinear local model training algorithm called nonlinear Bayesian weighted regression (NBWR) is proposed. In the NBWR, the nonlinear features of process data are first extracted by the autoencoder; then, given a query sample a local dataset is selected on the feature space and a fully Bayesian regression model with differentiated sample weights is developed. The benefits of this approach, which include better consistency of correlation, stronger abilities to deal with process nonlinearities and uncertainties, overfitting and numerical issues, lead to superior performance. The NBWR is used for developing a soft sensor under the LWL framework, and a real-world industrial process is used to evaluate the performance of the NBWR-based soft sensor. The experimental results demonstrate that the proposed method outperforms several benchmarking soft sensing approaches.

Stochastic Modeling of Groundwater Fluoride Contamination: Introducing Lazy Learners.

While it remains the primary source of safe drinking and irrigation water in northwest Iran's Maku Plain, the region's groundwater is prone to fluoride contamination. Accordingly, modeling techniques to accurately predict groundwater fluoride concentration are required. The current paper advances several novel data mining algorithms including Lazy learners [instance-based K-nearest neighbors (IBK); locally weighted learning (LWL); and KStar], a tree-based algorithm (M5P), and a meta classifier algorithm [regression by discretization (RBD)] to predict groundwater fluoride concentration. Drawing on several groundwater quality variables (e.g., concentrations), measured in each of 143 samples collected between 2004 and 2008, several models predicting groundwater fluoride concentrations were developed. The full dataset was divided into two subsets: 70% for model training (calibration) and 30% for model evaluation (validation). Models were validated using several statistical evaluation criteria and three visual evaluation approaches (i.e., scatter plots, Taylor and Violin diagrams). Although Na+ and Ca2+ showed the greatest positive and negative correlations with fluoride (r= 0.59 and -0.39, respectively), they were insufficient to reliably predict fluoride levels; therefore, other water quality variables, including those weakly correlated with fluoride, should be considered as inputs for fluoride prediction. The IBK model outperformed other models in fluoride contamination prediction, followed by KStar, RBD, M5P, and LWL. The RBD and M5P models were the least accurate in terms of predicting peaks in fluoride concentration values. Results of the current study can be used to support practical and sustainable management of water and groundwater resources.

Predicting the Enrollment and Dropout of Students in the Post-Graduation Degree using Machine Learning Classifier

Nowadays, In Bangladesh, the dropout rate at post-graduation level or incompletion of the post-graduation degree is considered as a serious problem in the education sector. This work can be used to support for identifying the specific individuals as well as the institutional factors which may next lead to the enrollment or drop out at the post-graduation degree. The real dataset is used to accomplish this work. Here, seven classification algorithms namely Naïve Bayes, Multilayer Perceptron, Logistic, Locally Weighted Learning (LWL), Random Forest, Random Tree, and Part are applied in this context. A confusion matrix is calculated for each classification model. Then, we computed all the seven performance evaluation metrics (accuracy, sensitivity, precision, specificity, F1 score, FPR, and FNR). Each classifier's performances are analyzed and measured from the computed performance evaluation metrics. Naïve Bayes, LWL, and Part classifier perform better than all other working classifiers attaining 86.36% accuracy and on the contrary, Random Tree classifier performs worst achieving 74.24% accuracy. After further analyzing of the result based on performance evaluation metrics, it is observed that LWL classifier performed best in this context among all the classifiers.

A Survey on Rice Crop Yield Prediction in India Using Improved Classification Technique

<p>India is an agricultural country. Agriculture is the important contributor to the Indian economy. There are many classification techniques like Support Vector Machine(SVM), LADTree, Natve Bayes, Bayesnet, K Nearest Neighbour(KNN), Locally Weighted Learning(LWL) on rice crop production datasets. They have some drawbacks like low accuracy and more errors. To achieve more significant result, To increase classification accuracy and reducing classification errors, our research uses classification method Bayesnet based adaboost will be proposed in work. Rice crop yield depend on environment's parameters like Rainfall, minimum temperature, average temperature, Maximum temperature, Vapour Pressure, potential evapotranspiration, reference crop evapotranspiration, cloud cover, wet day frequency for the kharif season. our dataset containing these environmental parameters for accurate prediction of Rice crop yield.</p>

Grid index subspace constructed locally weighted learning identification modeling for high dimensional ship maneuvering system

For off-line locally weighted learning (LWL), all training data points need to be stored in memory, which would lead to a heavy computational burden, especially for large amount of training data. To avoid heavy computational burden in LWL, the grid index subspace constructed algorithm is presented for high dimensional ship maneuvering system in this study. First, high dimensional training data can be encoded and stored in equal interval grid, and training data are divided into grids. Second, query point is encoded by using the same strategy as in the first step, and the grid number which belongs to the query point is obtained. Third, the subspace would be per-allocated to the query point by using the grid index which has a light computational complexity. Different from the general cluster algorithm, a subspace rather than a neighborhood is assigned to query point. This way, LWL is carried out in a subspace, and the computational complexity is significantly reduced. As a consequence, real-time performance is effectively guaranteed. Finally, theoretical calculations and simulation examples are given to validate the effectiveness of the proposed scheme.

A Comparative Study for Condition Monitoring on Wind Turbine Blade using Vibration Signals through Statistical Features: a Lazy Learning Approach

This study is to identify whether the wind turbine blades are in good or faulty conditions. If faulty, then the objective to find which fault condition are the blades subjected to. The problem identification is carried out by machine learning approach using vibration signals through statistical features. In this study, a three bladed wind turbine was chosen and faults like blade cracks, hub-blade loose connection, blade bend, pitch angle twist and blade erosion were considered. Here, the study is carried out in three phases namely, feature extraction, feature selection and feature classification. In phase 1, the required statistical features are extracted from the vibration signals which obtained from the wind turbine through accelerometer. In phase 2, the most dominating or the relevant feature is selected from the extracted features using J48 decision tree algorithm. In phase 3, the selected features are classified using machine learning classifiers namely, K-star (KS), locally weighted learning (LWL), nearest neighbour (NN), k-nearest neighbours (kNN), instance based K-nearest using log and Gaussian weight kernels (IBKLG) and lazy Bayesian rules classifier (LBRC). The results were compared with respect to the classification accuracy and the computational time of the classifier.

Multi-Innovation Gradient Iterative Locally Weighted Learning Identification for A Nonlinear Ship Maneuvering System

This paper explores a highly accurate identification modeling approach for the ship maneuvering motion with fullscale trial. A multi-innovation gradient iterative (MIGI) approach is proposed to optimize the distance metric of locally weighted learning (LWL), and a novel non-parametric modeling technique is developed for a nonlinear ship maneuvering system. This proposed method's advantages are as follows: first, it can avoid the unmodeled dynamics and multicollinearity inherent to the conventional parametric model; second, it eliminates the over-learning or underlearning and obtains the optimal distance metric; and third, the MIGI is not sensitive to the initial parameter value and requires less time during the training phase. These advantages result in a highly accurate mathematical modeling technique that can be conveniently implemented in applications. To verify the characteristics of this mathematical model, two examples are used as the model platforms to study the ship maneuvering.

Modified genetic optimization-based locally weighted learning identification modeling of ship maneuvering with full scale trial

This paper explores a novel nonparametric identification modeling technique for ship maneuvering system. In order to solve the over-learning or under-learning problem which may exist in distance metric optimization process in the locally weighted learning (LWL), a modified genetic algorithm (GA) is proposed. In our algorithm, a novel fitness function is defined, which can assign the maximum fitness to the optimal distance metric. And the performance analysis is developed based on the schema theory. Additionally, GA is a global search algorithm that the locally optimal is avoided. The LWL is applied to identify the characteristics of ship maneuvering motion by using the optimal distance metric. The proposed scheme improves the nonlinear mapping ability of LWL especially in the highly nonlinear area. The illustrative examples are utilized to demonstrate the effectiveness of the proposed scheme, including a synthetic data set and the YUKUN scientific research vessel of Dalian Maritime University. The simulation results indicate that the proposed scheme is a powerful modeling tool for ship maneuvering system.

Performance Exploratory Comparative Study for Software Change and Effort Prediction in Object-Oriented Systems Using Statistical and Machine Learning Algorithms

Precise software effort estimation is a vital sector of the software procedure. An excess as well as lessening software effort can prompt to unsafe concludes. Additionally, software venture supervisors need toward create evaluations of exactly how abundant a software development has been accepted cost. The predominant price of some software is the price of ascertaining effort. In this way, effort assessment has been exceptionally urgent as well as an effective dependably a necessity to enhance individual’s precision however as much at the time that could be expected. Different effort estimation models are, yet effortful to define and that model contributes too much precise assessment on that dataset. Above research experimentally assesses as well as analyzes the capability of Linear Regression Model, Bagging, Decision Table, Decision Stump, IBk, KStar, Locally weighted learning (LWL), Multilayer Perceptron, Simple Linear Regression, Stacking and Vote on software venture dataset. The dataset has been attained taken away from 3518 instances of java open source software. The outcomes demonstrate that Root relative squared error of K-Star technique was just 71.54%. Hence, the performance of K-Star technique has proved to be superior to the various other strategies. Above research, on the utilization of statistical and machine learning conclusion aimed at exploring the outcomes acquired in an analysis in the arena of computational intelligence has been concentrated. An investigation which includes a set of strategies in classification tasks has been shown and along with a set of procedures which are helpful to break down the conduct of a strategy as for a set of algorithms has been concentrated, for example, the structure in which another proposition is created.

Composite adaptive locally weighted learning control for multi-constraint nonlinear systems

A composite adaptive locally weighted learning (LWL) control approach is proposed for a class of uncertain nonlinear systems with system constraints, including state constraints and asymmetric control saturation in this paper. The system constraints are tackled by considering the control input as an extended state variable and introducing barrier Lyapunov functions (BLFs) into the backstepping procedure. The system uncertainty is approximated by a composite adaptive LWL neural networks (NNs), in which a prediction error is constructed by using a series-parallel identification model, and NN weights are updated by both the tracking error and the prediction error. The update law with composite error feedback improves uncertainty approximation accuracy and trajectory tracking accuracy. The feasibility and effectiveness of the proposed approach have been demonstrated by formal proof and simulation results.

153 - The Application of Machine Learning Algorithms to Immunophenotypic Data to Aid in the Diagnosis of MDS

Vapour phase detection of explosives using pattern recognition approaches is a very important area of research worldwide. This paper elaborates on the comparison between different algorithms in classifying empirical multiparametric data that are obtained from the explosive vapor sensors based on organic field effect transistors (OFETs). We address the problem of classification by means of statistical comparison among algorithms such as NaiveBayes (NBS), locally weighted learning (LWL), sequential minimal optimization (SMO) and J48 decision tree on data acquired from OFETs. This analysis helps in understanding the nature of data obtained from experiments and in making efficient estimators for the detection of explosives. The correctly classified instances for predicting tested samples using LWL, NBS, SMO and J48 decision tree are 72%, 73%, 80% and 90%, respectively. The future development of standoff explosive detectors will be benefited greatly by a proper choice of these classification approaches.

Comparison among different algorithms in classifying explosives using OFETs

Abstract Vapour phase detection of explosives using pattern recognition approaches is a very important area of research worldwide. This paper elaborates on the comparison between different algorithms in classifying empirical multiparametric data that are obtained from the explosive vapor sensors based on organic field effect transistors (OFETs). We address the problem of classification by means of statistical comparison among algorithms such as NaiveBayes (NBS), locally weighted learning (LWL), sequential minimal optimization (SMO) and J48 decision tree on data acquired from OFETs. This analysis helps in understanding the nature of data obtained from experiments and in making efficient estimators for the detection of explosives. The correctly classified instances for predicting tested samples using LWL, NBS, SMO and J48 decision tree are 72%, 73%, 80% and 90%, respectively. The future development of standoff explosive detectors will be benefited greatly by a proper choice of these classification approaches.

Scalable Techniques from Nonparametric Statistics for Real Time Robot Learning

Locally weighted learning (LWL) is a class of techniques from nonparametric statistics that provides useful representations and training algorithms for learning about complex phenomena during autonomous adaptive control of robotic systems. This paper introduces several LWL algorithms that have been tested successfully in real-time learning of complex robot tasks. We discuss two major classes of LWL, memory-based LWL and purely incremental LWL that does not need to remember any data explicitly. In contrast to the traditional belief that LWL methods cannot work well in high-dimensional spaces, we provide new algorithms that have been tested on up to 90 dimensional learning problems. The applicability of our LWL algorithms is demonstrated in various robot learning examples, including the learning of devil-sticking, pole-balancing by a humanoid robot arm, and inverse-dynamics learning for a seven and a 30 degree-of-freedom robot. In all these examples, the application of our statistical neural networks techniques allowed either faster or more accurate acquisition of motor control than classical control engineering.