Knowledge-based Model Research Articles

Knowledge Graph based Mutual Attention for Machine Reading Comprehension over Anti-Terrorism Corpus

ABSTRACT Machine reading comprehension has been a research focus in natural language processing and intelligence engineering. However, there is a lack of models and datasets for the MRC tasks in the anti-terrorism domain. Moreover, current research lacks the ability to embed accurate background knowledge and provide precise answers. To address these two problems, this paper first builds a text corpus and testbed that focuses on the anti-terrorism domain in a semi-automatic manner. Then, it proposes a knowledge-based machine reading comprehension model that fuses domain-related triples from a large-scale encyclopedic knowledge base to enhance the semantics of the text. To eliminate knowledge noise that could lead to semantic deviation, this paper uses a mixed mutual attention mechanism among questions, passages, and knowledge triples to select the most relevant triples before embedding their semantics into the sentences. Experiment results indicate that the proposed approach can achieve a 70.70% EM value and an 87.91% F1 score, with a 4.23% and 3.35% improvement over existing methods, respectively.

Knowledge-based mechanistic modeling accurately predicts disease progression with gefitinib in EGFR-mutant lung adenocarcinoma

Lung adenocarcinoma (LUAD) is associated with a low survival rate at advanced stages. Although the development of targeted therapies has improved outcomes in LUAD patients with identified and specific genetic alterations, such as activating mutations on the epidermal growth factor receptor gene (EGFR), the emergence of tumor resistance eventually occurs in all patients and this is driving the development of new therapies. In this paper, we present the In Silico EGFR-mutant LUAD (ISELA) model that links LUAD patients’ individual characteristics, including tumor genetic heterogeneity, to tumor size evolution and tumor progression over time under first generation EGFR tyrosine kinase inhibitor gefitinib. This translational mechanistic model gathers extensive knowledge on LUAD and was calibrated on multiple scales, including in vitro, human tumor xenograft mouse and human, reproducing more than 90% of the experimental data identified. Moreover, with 98.5% coverage and 99.4% negative logrank tests, the model accurately reproduced the time to progression from the Lux-Lung 7 clinical trial, which was unused in calibration, thus supporting the model high predictive value. This knowledge-based mechanistic model could be a valuable tool in the development of new therapies targeting EGFR-mutant LUAD as a foundation for the generation of synthetic control arms.

A combined scheme of parallel-reaction kinetic model and multi-layer artificial neural network model on pyrolysis of Reed Canary

A comprehensive understanding of pyrolysis kinetics is crucial for the design of biomass pyrolysis. In this study, a combined scheme was proposed for biomass pyrolysis. A kinetic model, serving as a knowledge-based model, was established to capture the primary trend of biomass pyrolysis. An ANN model, as an experience-based model, was developed to provide “details” on the process. The kinetic model incorporated a least-square optimization model to calculate optimal kinetic parameters for pyrolysis reaction steps. The ANN model utilized metaheuristic algorithms such as Particle Swarm Optimization, Pattern Search, Genetic Algorithms, and Surrogate Optimization to optimize the hyperparameters of the ANN model. By combining the kinetic model with the ANN model, comprehensive predictions of biomass pyrolysis were obtained, and the combined Kinetic-ANN model was validated with experimental data at 4 heating rates. Finally, the combined model was compared with the solo-kinetic and solo-ANN models, confirming the advanced performance of the combined approach.

Multi-view graph representation with similarity diffusion for general zero-shot learning

Zero-shot learning (ZSL) aims to predict unseen classes without using samples of these classes in model training. The ZSL has been widely used in many knowledge-based models and applications to predict various parameters, including categories, subjects, and anomalies, in different domains. Nonetheless, most existing ZSL methods require the pre-defined semantics or attributes of particular data environments. Therefore, these methods are difficult to be applied to general data environments, such as ImageNet and other real-world datasets and applications. Recent research has tried to use open knowledge to enhance the ZSL methods to adapt it to an open data environment. However, the performance of these methods is relatively low, namely the accuracy is normally below 10%, which is due to the inadequate semantics that can be used from open knowledge. Moreover, the latest methods suffer from a significant ”semantic gap” problem between the generated features of unseen classes and the real features of seen classes. To this end, this paper proposes a multi-view graph representation with a similarity diffusion model, applying the ZSL tasks to general data environments. This model applies a multi-view graph to enhance the semantics fully and proposes an innovative diffusion method to augment the graph representation. In addition, a feature diffusion method is proposed to augment the multi-view graph representation and bridge the semantic gap to realize zero-shot predicting. The results of numerous experiments in general data environments and on benchmark datasets show that the proposed method can achieve new state-of-the-art results in the field of general zero-shot learning. Furthermore, seven ablation studies analyze the effects of the settings and different modules of the proposed method on its performance in detail and prove the effectiveness of each module.

Data-informed reservoir computing for efficient time-series prediction

We propose a new approach to dynamical system forecasting called data-informed-reservoir computing (DI-RC) that, while solely being based on data, yields increased accuracy, reduced computational cost, and mitigates tedious hyper-parameter optimization of the reservoir computer (RC). Our DI-RC approach is based on the recently proposed hybrid setup where a knowledge-based model is combined with a machine learning prediction system, but it replaces the knowledge-based component by a data-driven model discovery technique. As a result, our approach can be chosen when a suitable knowledge-based model is not available. We demonstrate our approach using a delay-based RC as the machine learning component in conjunction with sparse identification of nonlinear dynamical systems for the data-driven model component. We test the performance on two example systems: the Lorenz system and the Kuramoto–Sivashinsky system. Our results indicate that our proposed technique can yield an improvement in the time-series forecasting capabilities compared with both approaches applied individually, while remaining computationally cheap. The benefit of our proposed approach, compared with pure RC, is most pronounced when the reservoir parameters are not optimized, thereby reducing the need for hyperparameter optimization.

Understanding Natural Language in Context

Recent years have seen an increasing number of applications that have a natural language interface, either in the form of chatbots or via personal assistants such as Alexa (Amazon), Google Assistant, Siri (Apple), and Cortana (Microsoft). To use these applications, a basic dialog between the assistant and the human is required. While this kind of dialog exists today mainly within static robots that do not make any movement in the household space, the challenge of reasoning about the information conveyed by the environment increases significantly when dealing with robots that can move and manipulate objects in our home environment. In this paper, we focus on cognitive robots, which have some knowledge-based models of the world and operate by reasoning and planning with this model. Thus, when the robot and the human communicate, there is already some formalism they can use -- the robot’s knowledge representation formalism. In this paper we describe an approach for translating natural language directives into the robot's formalism, allowing much more complicated household tasks to be completed. We do so by combining off-the-shelf SoTA large language models, planning tools, and the robot knowledge of the state of the world and of its own model. This results in much more accurate interpretation of directives in natural language.

Predicting Urban Heat Island severity on the census-tract level using Bayesian networks

Urban development and population growth have resulted in several critical impacts on the society, environment, and economy. One of the main impacts is the increase in temperature observed in urban areas, also known as the Urban Heat Island (UHI) effect. UHI has become the focus of several research studies due to the associated negative implications it causes. Despite that some research efforts examined different characteristics of the UHI phenomenon, there is a gap in the literature in terms of developing interpretable machine learning models that can accurately estimate or predict the severity of UHI (rather than air temperature or other UHI characteristics) as well as that can provide predictions on the census-tract level in the US. To that extent, this paper fills this gap by developing a knowledge-based white-box Bayesian network model that predicts UHI severity based on demographic, meteorological, and land use/land cover factors. First, a dataset for all census tracts in the state of New Jersey, USA was developed, which is comprised of 13 independent variables or factors affecting UHI severity. Second, expert knowledge was obtained from 10 UHI experts using the systematic three-round Delphi method to develop four different Bayesian network models. Third, the performance of the four Bayesian models was assessed and compared to choose the optimal model with the highest accuracy. Finally, sensitivity analysis was conducted to assess the influence of each key factor on the UHI severity. The results showed that the optimal model is a tree-augmented Bayesian network that can predict or estimate the UHI severity with an accuracy of 87.88%. The outcomes of this paper also reflected that the following 8 variables are the key factors that impact UHI severity: NDVI during winter season, NDVI during summer season, imperviousness, tree canopy, building area, population density, water body areas, and annual rainfall. The findings also identified the land use/land cover category as the major category affecting UHI severity compared to demographic- and meteorological-related factors. The proposed white-box Bayesian network model in this paper adds to the body of knowledge by allowing practitioners and researchers to perform micro-level UHI predictions and inferences about future or unknown UHI severity pertaining to individual communities or small/localized geographical areas. This enables more focused and targeted UHI mitigation actions to be planned, designed, and implemented in the most affected communities, which ultimately results in better decisions, actions, and outcomes to reduce or address the UHI effect.

An artificial neural network toward simulating the treatment of wastes in multiphase reactors by ozonation

The simulation of the performance of multiphase reactors is a difficult task with a high degree of uncertainties, due to the complex flow regimes, and multiple mass-transfer and reactive processes. A dimensionless artificial neural network (ANN) is developed to simulate the remediation of muds, heavily polluted with organic wastes. The mud is pretreated with synthetic seawater (SW) and anionic surfactant sodium dodecyl sulphate (SDS), then is placed inside a semi-batch bubble flow column, where the organic species are oxidized with the injection of an ozone-rich gas. The process performance is evaluated by measuring the removal efficiency of the total organic carbon (TOC) in dried mud. Experimental data are employed to calibrate a knowledge-based model (KBM), used as a tool for data regularization, and generation of simulated results to be used for ANN. A feed-forward dense multilayer ANN is developed, using a group of dimensionless variables as inputs and the KBM-simulated TOC removal efficiency in the liquid and solid phase of mud, as outputs. The interpretation of ANN results with the Shapley Additive explanation (SHAP) method shows that the TOC removal efficiency is enhanced when (i) the initial TOC concentration in solid phase increases; (ii) the mud is pre-treated with seawater/surfactant solution leading to the increase of the initial TOC concentration in liquid phase, and decrease of the initial TOC concentration in solid phase; (iii) the ozone concentration of injected gas increases; (iv) the Reynolds number in liquid phase takes on moderate values, facilitating the ozone dissolution, its reaction with dissolved TOC, and the mixing of phases; (v) the SDS concentration is quite high and, due to foaming, the apparent liquid column height, and the retention time of gas increase.

Knowledge-based BERT word embedding fine-tuning for emotion recognition

Emotion recognition has received considerable attention in recent years, with the popularity of social media. It is noted, however, that the state-of-the-art language models such as Bidirectional Encoder Representations from Transformers (BERT) may not produce the best performance in emotion recognition. We found the main cause of the problem is that the embedding of emotional words from the pre-trained BERT model may not exhibit high between-class difference and within-class similarity. While BERT model fine-tuning is a common practice when it is applied to specific tasks, this may not be practical in emotion recognition because most datasets are small and some texts are short and noisy, without containing much useful contextual information. In this paper, we propose to use the knowledge of emotion vocabulary to fine-tune embedding of emotional words. As a separate module independent of the embedding learning model, the fine-tuning model aims to produce emotional word embedding with improved within-class similarity and between-class difference. By combining the emotionally discriminative fine-tuned embedding with contextual information-rich embedding from pre-trained BERT model, the emotional features underlying the texts could be more effectively captured in the subsequent feature learning module, which in turn leads to improved emotion recognition performance. The knowledge-based word embedding fine-tuning model is tested on five datasets of emotion recognition, and the results and analysis demonstrate the effectiveness of the proposed method.

Resilient Formation Reconfiguration for Leader–Follower Multi-UAVs

Among existing studies on formation reconfiguration for multiple unmanned aerial vehicles (multi-UAVs), the majority are conducted on the assumption that the swarm scale is stationary. In fact, because of emergencies, such as communication malfunctions, physical destruction, and mission alteration, the scale of the multi-UAVs can fluctuate. In these cases, the achievements of formation reconfiguration for fixed-scale multi-UAVs are no longer applicable. As such, in this article, the formation reconfiguration problem of leader–follower multi-UAVs is investigated with a variable swarm scale taken into consideration. First, a streamlined topological structure is designed on the basis of the parity of the vertex numbers. Then, three formation reconfiguration strategies corresponding to the scenarios covering leader disengagement, follower detachment, and new member additions are developed with the aim of reducing the frequency of connection changes. Moreover, in terms of the leader election link of the leader disengagement scenario, a knowledge-based performance assessment model for UAVs is constructed with the help of the hierarchical belief rule base (BRB). Finally, the proposed formation reconfiguration strategies for leader disengagement, new member addition, and follower disengagement are demonstrated through simulations. The connection retention rate (CRR) for swarm communication topology under the three formation reconfiguration strategies can reach 67%, 90%, and 100%, respectively.

Boosting the prediction accuracy of a process-based greenhouse climate-tomato production model by particle filtering and deep learning

By generating high quality data without the big time investment and economic cost of real experiments, dynamic greenhouse climate and crop simulation models can support decisions on greenhouse climate control, crop management and greenhouse design. The reliability of simulation-based decisions depends on both the prediction accuracy and interpretability of simulation models. The prediction accuracy of these simulation models can be increased by: 1) improving mechanisms in process-based models; 2) calibrating process-based model parameters; 3) deriving black-box relationships from data. Considering the descending interpretability from (1) to (3), this study presents a knowledge-based data-driven modelling approach where firstly a process-based model is selected and modified based on domain knowledge, then data-driven improvement is applied including two steps: parameter value estimation by particle filter (PF) and further black-box improvement by deep neural networks (DNN). The approach was tested with an example of greenhouse climate-tomato production system modelling. Modules from GreenLight (Katzin et al., 2020) and TOMSIM (Heuvelink, 1995, 1996) were selected, modified and integrated into a process-based greenhouse climate-tomato model. Validation showed that PF-calibration of five greenhouse parameters decreased the seasonal relative root mean squared error (RRMSE) of indoor air vapor pressure predictions from 40.7% of that before PF-calibration to 16.4%, while it did not decrease the RRMSE of indoor air temperature predictions. Combining the PF-calibrated model with a DNN trained on a season of data decreased the RRMSE of indoor air temperature from 15.0% without DNN to 6.7%, and decreased the RRMSE of indoor air vapor pressure to 12.6%. The knowledge-based data-driven greenhouse climate-tomato model had a relative error of 0.9% for seasonal total fresh yield, and an RRMSE of 6.6% for the cumulative yield throughout the season. If process-based model parameters were not calibrated before combining the model with DNNs, the required amount and diversity of DNN training data increased because more information needed to be learnt from data by the DNNs. Without PF-calibration, combining a DNN trained on 50 days of data with the process-based model resulted in RRMSEs of 44.8% and 31.8% for indoor air temperature and vapor pressure prediction, respectively; with PF-calibration, the RRMSEs were decreased to 13.1% and 17.9%. The proposed three-step knowledge-based data-driven approach can not only improve the model prediction accuracy, but can also help to track and interpret the improvements.

Reaction performance prediction with an extrapolative and interpretable graph model based on chemical knowledge

Accurate prediction of reactivity and selectivity provides the desired guideline for synthetic development. Due to the high-dimensional relationship between molecular structure and synthetic function, it is challenging to achieve the predictive modelling of synthetic transformation with the required extrapolative ability and chemical interpretability. To meet the gap between the rich domain knowledge of chemistry and the advanced molecular graph model, herein we report a knowledge-based graph model that embeds the digitalized steric and electronic information. In addition, a molecular interaction module is developed to enable the learning of the synergistic influence of reaction components. In this study, we demonstrate that this knowledge-based graph model achieves excellent predictions of reaction yield and stereoselectivity, whose extrapolative ability is corroborated by additional scaffold-based data splittings and experimental verifications with new catalysts. Because of the embedding of local environment, the model allows the atomic level of interpretation of the steric and electronic influence on the overall synthetic performance, which serves as a useful guide for the molecular engineering towards the target synthetic function. This model offers an extrapolative and interpretable approach for reaction performance prediction, pointing out the importance of chemical knowledge-constrained reaction modelling for synthetic purpose.

Degree apprenticeships in the UK higher education institutions – are they viable?: Integrative literature review

PurposeApprenticeships for higher education institutes (HEIs) at degree and postgraduate levels in the UK have been challenging to provide, requiring increasingly complex collaboration and investment. This questions longer-term viability, creating a new perspective investigating evidence of value and impact for UK HEIs. This integrative literature review examines existing apprenticeship literature and showcases new knowledge for HEIs to update and inform future decision-making. Knowledge created offers a new framework, an “apprenticeship knowledge-based checklist model” and a perspective to explore further.Design/methodology/approachThis integrative literature review examines apprenticeship provision viability for UK HEIs across several disciplines since commencement.FindingsAnalysis of selected studies across three research themes critiques potential apprenticeship practice for HEIs: stakeholder collaboration, widening participation and work-integrated learning. Results provided a new integrated perspective via governance levels: strategic, academic department, and teaching team.Research limitations/implicationsThe research draws from multiple disciplines representing current literature in the HEI Apprenticeships field. However, the context in which programmes are delivered by HEIs and employers differs significantly, so factors continue to emerge.Practical implicationsPractical implications drawn from these findings aim to cultivate dialogue before embarking/withdrawing from HEI Apprenticeship programmes. It offers employers guidance expectations when partnering with university providers. For global apprenticeship provisions, it offers a checklist for re-valuating practice.Originality/valueThe research identified new knowledge implications from existing literature in a field experiencing rapid growth. It offers valuable contributions: a knowledge-based checklist model constructed for HEIs to evaluate existing and future viability provisions at each governance level. It is a catalyst for new practice approaches and an agenda for further research.

Knowledge-based and data-driven behavioral modeling techniques in engagement simulation

As knowledge and data increase in scale and complexity, it is more difficult to apply these two key assets to achieve optimal effectiveness in engagement simulation. The aim of this study was to investigate the techniques of knowledge and data integration with respect to the development of smart agents to predict accurate behaviors in tactical engagements. To reduce the complexity of combat behavior representation, with respect to the functions, we represented subject matter expert operational knowledge by proposing multiple levels of cascaded hierarchical structure, namely, the function decision tree, to increase the readability and maintainability of the behavioral model. For decision points in a behavioral model, smart agents can be trained based on data samples collected from rounds of constructive simulations which provide validated physical models and tactical principles. As a proof of concept, we constructed a simulation testbed of multi-warhead ballistic missile penetration, which generated 129,600 constructive simulations over a total of 84 h. Thereafter, we selected 5817 data samples (i.e. ~4.5% of the simulations) using an operational metric of total rewards exceeding 100. The data samples are used to train an artificial neural network and then this network is used to develop a deep reinforcement learning agent. The results revealed that the training process iterated nearly 17,000 epochs until the policy loss decreased to an acceptable low value. The smart agent increased the ratio of ballistic missile target hits by 18.96%, a significant increase when compared with the traditional rule-based behavioral model.

Diagnosing attention-deficit hyperactivity disorder (ADHD) using artificial intelligence: a clinical study in the UK.

Attention-deficit hyperactivity disorder (ADHD) is a neurodevelopmental disorder affecting a large percentage of the adult population. A series of ongoing efforts has led to the development of a hybrid AI algorithm (a combination of a machine learning model and a knowledge-based model) for assisting adult ADHD diagnosis, and its clinical trial currently operating in the largest National Health Service (NHS) for adults with ADHD in the UK. Most recently, more data was made available that has lead to a total collection of 501 anonymized records as of 2022 July. This prompted the ongoing research to carefully examine the model by retraining and optimizing the machine learning algorithm in order to update the model with better generalization capability. Based on the large data collection so far, this paper also pilots a study to examine the effectiveness of variables other than the Diagnostic Interview for ADHD in adults (DIVA) assessment, which adds considerable cost in the screenining process as it relies on specially trained senior clinicians. Results reported in this paper demonstrate that the newly trained machine learning model reaches an accuracy of 75.03% when all features are used; the hybrid model obtains an accuracy of 93.61%. Exceeding what clinical experts expected in the absence of DIVA, achieving an accuracy of 65.27% using a rule-based machine learning model alone encourages the development of a cost effective model in the future.

Implementation of a knowledge-based mechanistic model of EGFR-mutant lung adenocarcinoma that considers tumor heterogeneity and metastases to reproduce disease progression upon EGFR-TKI administration.

e21190 Background: 16,4% of lung adenocarcinomas (LUAD) are presenting a mutation in the epidermal growth factor receptor (EGFR), resulting in its constitutive activation and leading to uncontrolled cell proliferation [1]. Tyrosine kinase inhibitors (TKI) have been developed to inhibit EGFR activity but the presence of metastases and resistance mutations explain the lack of durable response to the treatment [2]. Knowledge-based mechanistic models reproducing existing clinical trials, based on population characteristics, can be used to help the design of future clinical trials. In particular, they can inform on the best responders to given treatments. Methods: We developed a physiologically based pharmacokinetic model of osimertinib, a 3rd generation TKI, to account for the distribution of the drug in the primary tumor and metastases after oral administration. This model was then combined to a pathophysiological mechanistic model of EGFR-mutant LUAD to represent the impact of osimertinib on the signals arising from EGFR activation. The combined model outputs the evolution of the primary tumor and each metastasis to allow the evaluation of the patient progression according to the RECIST criteria. Furthermore, each tumor in the model is composed of several subclones each possessing their own set of mutations and therefore responding differently to the treatment. Data from the clinical trials FLAURA and AURA3, in which osimertinib was given respectively as first and second line, were used to calibrate the model. Visual predictive checks as well as statistical tests were performed to ensure the proper behavior of the model. Results: The model successfully reproduced the time to progression in an EGFR mutant LUAD population treated with osimertinib as first line or as second line. In addition, the model reproduced the causes of progression according to the RECIST criteria and the sites of apparition of new metastases (in lung, brain, liver and bone). Conclusions: Reproduction of real world data brings credibility to the model. This is essential to use the model as an investigational tool to provide relevant insights, potentially on the best responders to osimertinib. After validation with additional clinical patient level data, the model could be used to create synthetic control arms in upcoming clinical trials. This would grant an improved analysis of covariate relationships with the comparison of an investigational treatment to the standard of care osimertinib administered as first or second line. It would also reduce the number of patients needed in the trial.

Genetic algorithm based production knowledge base for mechanical fault detection model

Mechanical fault detection has an important influence on production schedule and efficiency. With the development of intelligent technology, more and more intelligent detection technologies are applied to mechanical fault detection. In order to detect mechanical faults more efficiently and accurately, this experiment proposes a production knowledge base model based on genetic algorithm (GA algorithm). The model uses the unique biological genetics principle of genetic algorithm to evolve the interested population, and can conduct spatial search to find the global optimal solution. By comparing the performance of GA algorithm model with other similar detection models, it is found that the model proposed in the experiment has obvious advantages in mechanical fault detection performance. The experimental results show that the maximum accuracy of the GA algorithm is 0.935, 0.074 higher than the support vector machine (SVM) model, 0.118 higher than the linear discriminant analysis (LDA) model, 0.032 higher than the random forest (RF) model, and 0.166 higher than the K nearest neighbor (KNN) model. In addition, the error value of GA algorithm is the lowest among these models, which is 0.028. This proves that the genetic algorithm model has higher diagnostic accuracy and can play an important role in mechanical fault detection.

Integration of the Wang & Mendel Algorithm into the Application of Fuzzy Expert Systems to Intelligent Clinical Decision Support Systems

The use of intelligent systems in clinical diagnostics has evolved, integrating statistical learning and knowledge-based representation models. Two recent works propose the identification of risk factors for the diagnosis of obstructive sleep apnea (OSA). The first uses statistical learning to identify indicators associated with different levels of the apnea-hypopnea index (AHI). The second paper combines statistical and symbolic inference approaches to obtain risk indicators (Statistical Risk and Symbolic Risk) for a given AHI level. Based on this, in this paper we propose a new intelligent system that considers different AHI levels and generates risk pairs for each level. A learning-based model generates Statistical Risks based on objective patient data, while a cascade of fuzzy expert systems determines a Symbolic Risk using symptom data from patient interviews. The aggregation of risk pairs at each level involves a fuzzy expert system with automatically generated fuzzy rules using the Wang-Mendel algorithm. This aggregation produces an Apnea Risk indicator for each AHI level, allowing discrimination between OSA and non-OSA cases, along with appropriate recommendations. This approach improves variability, usefulness, and interpretability, increasing the reliability of the system. Initial tests on data from 4400 patients yielded AUC values of 0.74–0.88, demonstrating the potential benefits of the proposed intelligent system architecture.

A Comparison of Two Machine Learning Techniques for the Prediction of Initial Oil in Place in the Niger Delta Region

Conventionally, the knowledge of experts on the drilling features of a potential oil well is practically used to predict the volume of initial oil in place. Experts used different knowledge-based models such as volumetric, material balancing, analogy to predict the initial oil in place. In this study, 816 datasets were collected from Shell petroleum development company (SPDC) where the volumetric method is used for their prediction. These datasets were preprocessed and applied on two machine learning techniques of random forest and supervised vector regressor to predict the initial oil in place and the results obtained were compared with that obtained from SPDC.The results of computation using 4 principal features from the 9 features were closer to that obtained from SPDC than the computations using all the 9 features. The results of computations with random forest were also compared with that of supervised vector regressor. The results of random forest covary strongly (0.970) with the field results more than that of the support vector regressor (0.832). The uniqueness of this study is shown in the use of 4 predicting features (independent variables) to obtain prediction values that are very close to that obtained in the field with 9 features. This is obtained with random forest, so it can be recommended as a reliable machine technique for the prediction of initial oil in place in the Niger delta region.

Resampling Technique for Imbalanced Class Handling on Educational Dataset

Educational data mining is an emerging field in data mining. The need for accurate in identifying student accomplishment on a course or maybe an upcoming course can help the institution to build technology-aided education better. Educational data mining becoming a more important field to be studied because of its potential to produce a knowledge base model to help even the teacher or lecturer. Like another classification task, educational data mining has a common and frequently discovered problem. The problem that occurred in educational data mining specifically and classification tasks generally is an imbalanced class problem. An imbalanced class is a condition where the distribution of each class is not in the same proportion. In this research, it is found that the class distribution is severely imbalanced and it is a multiclass dataset that consists of more than two class labels. According to the problem stated beforehand, this paper will focus on the imbalanced class handling and classification with several methods on both of it such as Linear Regression, Random Forest and Stacking for classification and SMOTE, ADASYN, and SMOTE-ENN for the resampling algorithm. The methods are being evaluated using a 10-fold cross-validation and an 80-20 splitting ratio. The result shows that the best performance coming from the Stacking classification on ADASYN resampled dataset evaluated using an 80-20 splitting ratio with a 0.97 F1 score. The result of this study also shows that the resampling technique improves classification performance. Even though the no-resampling classification result produced a decent result too, it can be caused by several things such as the general pattern of the data for each class is already been good from the start. Thus, there is no real drawbacks if the original data is processed.