COGNITIVE MACHINE LEARNING

Cognitive computing is used to explain AI systems that intend to pretend human cognition [13,18]. Many artificial intelligent technologies require a computer system. To build cognitive models that imitate social cognition processes, including ML, deep learning, neural networks (NNs), neurolinguistic programming, and sentiment analysis. This chapter represents the CoML framework for the learner. Using the CoML framework, the learners can elate the competency to predict their scope to embrace course. According to bloom’s technology, the cognitive skills of a learner are diversified into creating, evaluating analyzing, applying, understanding, and remembering [25]. ML techniques have been extensively investigated and enriched, hi supervised learning, the past experienced data with labels are processed scientifically. The perception, logistic regression, and the nearest-neighbor rule are new representative ML methods. Based on the perception, NNs and support vector machine (SVM) were proposed later. The ML models, Naive Bayes (NB), and Random Forest (RF), have been utilized more extended time and have shown more significant achievement beyond the diversity of errands. Unlike supervised ML, unsupervised learning is based on training data without labels. In unsupervised learning, data are taken as input and generate a group of data or cluster with the data records. The cluster analysis method is used for elementary data analysis to create trends and patterns in the group of data.

2.5 STATE OF THE ART

The literature survey was carried out with the published work hi Cognitive Neuroscience, Cognitive Computing, and CoML research domains. Table 2.1 presents a set of 25 relevant research papers. Colt et al. [26] have worked on a method for computing the scores of the key performance indicators for IT professionals to find the competency assessment process. They have estimated vital performance indicators by considering four performance levels. Limitation of this research is a small size of the sample as they have cited. They have implemented no computerized techniques. They have focused on problem-solving skills. The authors have identified this skill as essential for promoting graduates in the job. The reason is that during training, the skills that are acquired by the graduates and that employers look into candidates are problem-solving skills. The authors have presented a model to evaluate graduates with their problem-solving ability to meet with industry job competency. To train the proposed model, they have used ML techniques to predict suitable jobs for graduates. The authors have addressed challenges, which are faced by graduates and industry in the recruitment process. In their work, they have highlighted trends, methods, and gaps hi the recruitment process for skill assessment and expectation [27]. Kart [25] worked on learning-style detection-based cognitive skills. He focused on three objectives: the first is identifying the learning style based on the cognitive abilities of a novice vigorously, the second is a mapping between cognitive skills and Bloom’s taxonomy with learning object, and the third is deriving the knowledge competency level for improvement by way of keeping track and feedback mechanism through the reinforcement model. The implemented system classifies learners into two groups based on memory and concentration. Serby et al. [28] have worked to determine the efficiency in a postgraduate thoracoscopy program. In evaluating a single group using the test, the model comprises multiple-choice questions (MCQs) and psychomotor skill measures. They have taken pre- and postcourse training assessment data of the trainer to evaluate them. This work is adopted by other areas of medical educators to do procedural training based on the competency-based paradigm. Dlamini and Leung [29] worked to predict sales performance while hiring using PMaps scientific sale assessments. In their assessment tools, all participants who scored high in PMaps Sales Assessment™ have performed better at insurance sales job. They are going to extend their work to develop a predictive algorithm for the client for predicting likely sales performance of a candidate at the time of the prehiring stage. Van Lieshout et al. [30] have implemented ML techniques such as NB classifier, logical reasoning, SVM, decision tree method RF, and NNs. They have mentioned that the study of ML applied to email spam classification for a long time requires a tremendous amount of records for immense precision [31]. Their models accomplished superior performance with small and biased samples in comparison with other representative ML methods.

Currently, machine learning algorithms are used for predicting academic achievements. The ML-based app is used for self-evaluation of teacher- specific instructional methods and tools [31,32]. Most of the researchers have used ML techniques to predict student’s academic performance. Navalyal and Gavas [27] have compared classification techniques such as decision tree, LASSO, K-NN, and SVM to evaluate student’s performance. Authors of [34] have also used decision tree methods such as ID3, C4.5, and classification and regression tree (CART) techniques. To access the academic achievements of undergraduate students with qualitative attributes such as economic status, resource utilization, and living location. The authors of [35] have used CART decision tree with cross-validation and RF to evaluate student’s performance for engineering students. Delorme et al. [24] have developed intelligent tutorial systems, which promised to deliver an adaptive learning experience to improve student learning outcomes. The prospective of neuroscience training packages is to surge the inclusive brain utility affected by the neurocognitive functions. Taking abode in the brain circuits while performing core tasks such as reading. The mathematical solution was paving the way to augment the academic skill acquisition [25]. The consequence of using brain imaging methods is to seive as analytical tools for measuring the educational interferences of the learner. To read through the brain and to detect signals, noninvasive methods such as fMRI and EEG are used [26]. These neuroimaging methods have latent ability to enhance the cognitive subprocess at a profound level than reliance on only behavioral methods. The academic competence fosters better if neural and cognitive bases are considered. There is a strong connotation of imaging data with verbal strategy reports than with the problem size, validating the capture of brain activation in the process of mental tasks [27]. These insights can be applied to educational settings to measure the skills taught in the classroom, such as responding to reasoning and problem solving. Neuroeducation is a promising area spanning the gap to progress bidirectional activity in the brain of learners and facilitators impacting the learning process. The causal cognitive functions designed by learning environment confine the flexibility of brain circuits rendering to poor performance in academics. The neural science and cognitive bases of academic skills should be organized for shaping a better arrangement for learning environments that optimally achieve these capabilities according to the learning ability [28]. BCIs have not attained much potential proven outside the laboratory because of their robustness. The authors have followed the instructional design principles at several levels to identify the flaws in BCI training protocols. Here, the instructions are provided to the user in the tasks to execute and feedback provided to them. Poor performance is due to signal processing algorithms for analysis of EEG signals [34].

PROBLEM IDENTIFICATION

Based on the literature review, it has been found that in higher education having a competitive and a vast domain of knowledge attainment process, when the learner has to endorse into a specified course there is much ambiguity about the prerequisites ought to be acquired. The learner accommodates to the class as per the individual interest. Learner incompetency to obtain the competition results in end of the journey. The requirement, the cognitive data of the student, is to acquire in counting with the performance indicator to build a CoML model. Cognizance factors of shedding light for self- assessment are not a priori considered in addition to the course endorsement competency factors according to the existing state of the art. The research objective is to investigate whether there is a significant effect on students in activity-based teaching. To clarify the concept and their academic performance, as well as in the evaluation of the effectiveness of teaching methods, three supervised ML techniques are compared to find the accurate analytical result. The significance of this study provides better insight into the teaching community for the use of an effective teaching method used by them with ML techniques.

SETTING BENCHMARK

The proposed model is based on three areas of the existing system. The first area is cognitive theory, where the researchers have marked different angles of cognition, such as creativity, analyzing, comprehension, evaluation, and understanding. These functions take place in the brain. Many of the researchers have worked to associate the actions taken place in the mind to the cognition tasks. For reading a brain signal, a hardware device that can read the analog signals and convert them to the computer-readable format is required. This task is to be performed by the BCI. Reading the signs to prototype a model and then establish an ML algorithm to prove the accuracy of correctly classified instances is the domain of ML. This chapter compares the three areas, as mentioned in Table 2.1, which describes the work done by different authors implementing the technologies related to cognitive theory, BCI, and ML techniques applied to cognition. Many of the authors have involved one or more technologies. Competency assessment using key performance is applied for measuring IT professional competency but has not implemented the BCI. The CRESST model is used for learning cognitive theory for training and evaluation. The Kraiger model used AI and SVM for predictive mapping of graduate skills to industry roles (Mwakondo, 2016). Student’s academic performance prediction has been done using SYM [7]; competency-based metrics are used to determine the effectiveness of a postgraduate thoracoscopy course. Assessment is based on orientation style, personality, attitude, interest stream, and subject selection; career planning made EEG oscillations for memory retention [19]. The interplay of interests between natural and artificial cognitive systems for the spectacular medium. A tool to make disabled people communicate was developed using the EMOTIV instrument to measure EEG signals [20].

 
Source
< Prev   CONTENTS   Source   Next >