A Study on Troubleshooting Dynamic Investigation Based on Reasoning Techniques by Artificial Intelligence Framework

There has been noteworthy advancement in the improvement of programming and equipment for the investigation and design exercises in power system arranging, activity and control. Be that as it may, much still relies upon the judgment of human experts; i.e. experienced arranging and design personnel equipped for settling on instinctive and proficient choices based on the complete information of the common conditions. Development of the power system, advancement of innovation, high hardware dependability and the subsequent deficiency of involvement in managing flaws have caused worry in the business that the expertise in the arranging and task of power systems might be lost as staff retirements rise and substitutions are impeded. It is conceivable that artificial intelligence (AI) tools can fill this void between the requirement for, and the accessibility of, experts later on. Expert systems are powerful AI tools utilizing which the information of experts can be generally scattered. The constant progress in technology has not recently conveyed and a lot of new gadgets and relapsed enlisting power into various parts of our regular daily existences; it is in like manner driving a change on how society relates to Computer Science. The downsizing procedure in contraptions has formally made open a broad assortment of embedded enlisting gadgets which would now have the capacity to help us when we wash pieces of clothing and dishes, cook our meals, and drive our automobiles. These investigation advancements are immediately abused by overall undertakings, see for example, through the method of data trade, propelling 'globalization' of technology.

Intelligence is defined as the ability to acquire, retrieve, and use knowledge in a meaningful way, including both, raw and refined knowledge. It also includes the ability to memorize information, recall facts, and express emotions. The important aspects of human intelligence are:- • Judgment and decision making, • Commonsense, • Use of intuition to make choice of action about the forthcoming events, • Problem solving using creativity, • Goal directedness and planning, Despite the fact that the human intelligence is powerful, it has a few constraints. Humans are mentally error prone; they have constrained learning bases, and data handling of a sequential sort which continues all around gradually in mind when contrasted with that of PCs. Along these lines, the significance of intelligence isn't human cerebrum's data handling capacity yet the capacity of humans to exhibit their intelligence by conveying successfully, and by learning. Humans can secure learning by involvement and after that illustrate, by imparting the information that they have procured. Standard reasoning in psychology thinks about human intelligence as a capacity in-intellectual procedure. The present research in Artificial Intelligence is centered around the accompanying segments of intelligence:- Learning, Reasoning, Understanding, Creativity and Intuition. Looks into around the globe are planning to create PC systems displaying the above highlights and have been effective to certain degree as well.

At first, scientist's conviction that making an AI would disentangle recorded as a hard copy programs for every single intelligence work that performs. Experiencing with this assignment, it went to the acknowledgment that this methodology was excessively shallow. Indeed, even straightforward capacities like face acknowledgment, unique sense, and example acknowledgment and language appreciation were past their programming skills. It arrived at the resolution that making an AI needs a lot further learning into natural intelligence first. It is, in this way, required to see how perception, cognizance, basic leadership occur in the human personality and what seeing truly implies. In this manner, specialists spread into various approaches, some went into the investigation of the cerebrum and endeavored to see how the system of neurons makes the psyche yet they had a similar objective of making shrewd machines • The top down approach • The bottom up approach

Yang Xiao et al (2007) [1] introduced the utilization and future possibilities of telemedicine technology with models. A short exchange about LINCOS venture was additionally displayed. A point by point Medical Records EMRs which were viewed as the eventual fate of telemedicine systems was additionally informed upon. Mehdi Kashei et al (2012)[2]proposed a hybrid model of artificial neural networks utilizing multiple straight relapse models so as to get progressively exact classification precision. The model could be utilized for a 2 class and a multi class issue. Bagher Ebadian et al (2009) [3] proposed a technique to think about the ischemic stroke utilizing Artificial Neural Network (ANN). The strategy recognizes the degree of ischemic sore recuperation. Examinations utilizing new dataset demonstrated that the forecast made by the ANN had an incredible generally speaking execution and was all around corresponded to the 3-month ischemic injury on T2-Weighted image. Pinho (2013) [4] presented another strategy for digital image compression that consolidated two non-customary techniques to be specific, versatile block size image division and artificial neural networks. An algorithm for image compression utilizing these techniques was proposed and some test results were displayed. All compression parameters (for example the output values of the concealed units) were consistently and scalar quantized to 6 bits. The examination likewise focused on that, attributable to the versatile block size ability diminished the quantity of parameters to actualize compression system. El-Sayed et al (2009) [5] exhibited two hybrid techniques for the classification of the attractive reverberation human cerebrum images. The proposed hybrid strategy comprises of three phases, in particular, highlight extraction, dimensionality decrease, and classification. In the primary stage, features related with MRI images utilizing discrete wavelet transformation (DWT) were acquired. In the second stage, the features of attractive reverberation images (MRI) were diminished utilizing principles component analysis (PCA) to the more fundamental features. In the classification organize, two classifiers dependent on supervised machine learning were created. The principal classifier dependent on feed forward back-proliferation artificial neural system (FP-ANN) and the second classifier dependent on k-nearest neighbor (k-NN). The classifiers were utilized to classify subjects as typical or unusual MRI human images. A classification with an accomplishment of 95.6% and 98.6% was gotten by the two proposed classifiers FP-ANN and k-NN individually. Bainbridge, et al. (2014) [6] related the two boundaries of Social intelligence and Artificial

for hypothesis and research that may be accomplished by Artificial intelligence technology, Symbolic processors, expert system, neutral networks, hereditary algorithms and classifier systems. The primary significant achievements of Artificial-Social Intelligence (ASI) have been in the realm of hypothesis, where these techniques have enlivened new speculations just as rendering existing hypotheses increasingly thorough. Two application regions for which ASI holds incredible guarantee are the sociological analysis of composed tests and information retrieval from the inevitable Global Information Infrastructure. ASI has just been connected to certain sorts of factual analysis, however how focused it will be with increasingly conventional techniques stays indistinct. Woolsey and Johnson (2011) [7]characterize Naturalistic Intelligence, a great military activities inquire about analysis and very much doomed the most recent sausage specialized equipment or software contraption. Expert systems and Artificial intelligence and their purveyors and lackeys and treated with the entertainment they merit. Much more terrible we should not be so made up for lost time in the discussion that specific TTs are useless that we stop to be worried about things, and rather moved toward becoming concerned just with the best approach to talk about things, enabling science to be debased from a journey for truth to a verbal exercise. George F Ludger (2009) [8] who portrayed structures and procedures for artificial intelligence. The audit likewise contains the methodologies of artificial intelligence, for example, feeble artificial intelligence and solid artificial intelligence. It likewise thinks about the present real world applications and current procedures in artificial intelligence. Dwindle Norvig and Stuart Russell (2016) [9] in their examination paper "Artificial intelligence: A cutting edge approach" have actualized the fundamental points of view of artificial intelligence. They inferred that the artificial intelligence is a mix of reasoning, learning, perception, linguistic methodology and critical thinking. It likewise have depicted about the supervised kind of machine learning .The desire of this survey is to present the sorts of machine learning, for example, supervised, unsupervised and support and so on. The audit likewise investigates the uses of AI and machine learning in real time. William and Allen (2016) [10]Knowledge-based systems (KBS) are PC based systems which support, or perform naturally, psychological errands in a limited problem space which are typically just done by human specialists. KBS are not quite the same as conventional PC applications in various perspectives. Initial, a KBS speaks to knowledge actualize knowledge certainly as strategies and therefore can just apply it in a foreordained manner. Furthermore they are difficult to keep up as an update of the information is reflected by the alteration of a strategy. A KBS gives problem solving capacity which is performed by an inferential motor. It is likewise ready to legitimize its conduct by communicating the inferential advances that have prompted a specific end. The knowledge-based systems are characterized for the most part in seven kinds, they are: master systems, neural systems, case-based thinking, genetic calculations, wise operators, data mining and insightful mentoring systems. McCarty (2007)[11]Artificial intelligence and Law (AI and Law) is a sub-area of artificial intelligence (AI) for the most part worried about utilizations of AI to lawful informatics problems. The primary research worry of the field is to contribute in the utilization of most recent technological tools and systems accessible to the bigger endeavor of AI, there by bringing about mechanization of the legitimate informatics space. The space of AI and Law is in presence since L. Thorne McCarty's Harvard Law Review article, "Reflections on Taxman: An Experiment in Artificial Intelligence and Legal Reasoning". The field of artificial intelligence and law is conceived as an utilization of the bigger order of artificial intelligence. Artificial intelligence, from its absolute starting point, has favored the legitimate area for its applications. The exploration was especially roused by logic fitting which mixed in old philosophical custom to ground legitimate thinking in logical establishments. The thought is that lawful standards and thinking are coordinated in a system (Knowledge-based systems) which would enable human creatures to decide. This vision isn't without problems as it requires models of law that are calculable and diminish legitimate exercises in order to rules, to formalize law.

1. To test the attainability of a combined reasoning way to deal with a who dunnit problem known as Aunt Agatha 2. To use a combined reasoning procedure in constraint solving 3. To study better understanding of Artificial Intelligence

As mentioned above, we aim to employ different approaches (machine learning, constraint solving of using three different types of solvers to attack the same problem, we looked at the ―Who Killed Aunt Agatha‖ problem from the TPTP library (i.e., problem PUZ001, originally from). We cherry-picked a ―whodunnit‖ puzzle because it narrates a murder story in which the mystery is revealed by identifying the murderer. Such problems are akin to real life investigation scenarios we aim to tackle. Note that TPTP library contains a set of test problems for automated theorem proving (ATP) systems. The background knowledge for this problem, stated in English, as follows: ―Someone who lives in Dreadbury Mansion killed Aunt Agatha. Agatha, Butler and Charles live in Dreadbury Mansion and are the only people who live therein. A killer always hates the victim and is never richer than the victim. Charles hates no one that Aunt Agatha hates. Agatha hates everyone except the Butler. The Butler hates everyone not richer than Agatha. The Butler hates everyone Aunt Agatha hates. No one hates everyone and Agatha is not the Butler‖.

To show that – in principle – investigation problems are amenable to machine learning, we first tried to tackle Aunt Agatha problem by applying a machine learning methodology. We invented some data which embodies the axioms of the problem. In particular, we wrote down the details of five case studies with three people in, one of whom had been murdered.

We saw how to create investigation problems in the beginning times of the advancement of GH. These problems are referred to GH as current cases as GH accept every investigation problem as a present issue, and expects to tackle them by utilizing diverse AI procedures. As we clarified in section 2, by generating the past cases, we expect to utilize different procedures so as to become familiar with the profile of the liable party. This profile will be connected to the present case so as to take care of the problem or to make the problem simpler by featuring the bystanders or taking out a portion of the suspects. Reviewing from part 3, GH utilizes a similar age calculation with slight alterations for delivering past cases. Changes were made to abstain from problem: A patient is admitted to an emergency clinic upon coming back from Siberia. His side effects incorporate fever, rash, endless bone torment and hypertension. The specialist realizes that it is extremely regular to contract Pneumonia in chilly climate and pneumonia clarifies two of the manifestations. In any case, another ailment called Polio clarifies three of the manifestations. It is judicious to look at both of the ailments, yet the patient's condition is falling apart and the specialist can't bear the cost of the time required by various tests. A savvy specialist ought to dependably organize his/her judgment.

DIPs are dynamic as in they change at different time steps. Over the long haul, more proof is gathered from the present case and more information becomes exposed from the past cases. GH can search for a solution at each time step, which we call a transient solution, and eventually, the last solution is gotten when no more information is bolstered to GH. In our experiments, we assessed GH's fleeting solutions so as to measure GH's presentation at various time steps. For example, figuring the normal error rates when 20% of the information (suspects, facts and rules) is accessible. We should take note of that DIP types, as clarified in page 170, are elaborated into constituent factors of the embedded rules: various literals conjoined to shape the body of the ruleset(s) and (b) the most extreme arity of those literals. So as to think about GH's solution execution, we expected to ensure that all DIPs are assessed on a reasonable premise. Therefore, we designed GH to similarly circulate the urgent information among time intervals. GH generates the present case so that at each time step a piece of the embedded rule(s) is uncovered. In synopsis, the arrangement of literals showing up in the rule sets is separated into equivalent subsets where every subset will be exhibited at each time step. Reviewing from the rule set properties appeared in figure 5.3, Pi = {p1, p2, . . . , pk} is the arrangement of predicates showing up in the body of the rule Ci . In our dynamic experiments, at each time step, GH uncovers one of the predicates showing up in the Pi. Therefore, every one of the DIPs with rules that contain n predicates, are assessed in n time intervals. Give us a chance to consider a precedent appeared in figure 5.5. The figure portrays that the embedded rule contains predicates 4, 5 and 5. In that capacity, these 3 predicates ought to show up in 3 time intervals. Plainly pred4 shows up in t = 1

intervals. Along these lines, 33% of the embedded rule is constantly given to the GH solution module at every interim.

AI problems which models somewhat conventional circumstances which may emerge in, state, medical determination or the solving of a wrongdoing. We utilized the term investigation problems to mean such problem where the objective is to pick a guilty gathering, in charge of a marvel happening in a medical or criminal investigation, from various suspects. That is, there are various potential judgments/suspects (candidates), and the problem is to utilize the facts of the case to rank them arranged by improving probability of being the reason for the illness/guilty of the wrongdoing (which we call the target candidate). Such positioning regularly prompts further medical tests/police enquiries concentrating on the probably candidates, which will uncover further information about the present case. Consequently, we utilize the term dynamic investigation problem (DIP) to depict a progression of such problems to be settled. Solving every problem involves utilizing the facts of the case, coupled with earlier knowledge about the area to limit the candidates to only one.

[1]. Yang Xiao, Daisuke Takahashi and Fei Hu (2007). ―Telemedicine Usage and Potentials‖, IEEE WNC Proceedings, pp. 2738-2742. [2]. Mehdi Khashei, Ali Zeinal Hamadani, Mehdi Bijari (2012). ―A novel hybrid classification model of artificial neural networks and multiple linear regression models‖, Expert Systems with Applications 39, pp. 2606–2620. [3]. Bagher-Ebadian H., Jafari-Khouzani K., Mitsias P.D., Soltanian Zadeh H., Chopp M., Ewing J.R. (2009). ―Predicting final extent of ischemic infarction using an artificial neural network analysis of multiparametric MRI in patients with stroke‖, International Joint Conference on Neural Networks, pp. 229-235. [4]. Pinho J. (2013). ―Image Compression Based on Quadtree Segmentation and Artificial neural networks‖, Electronics Letters, Vol. 29, No. 11, pp. 1029-103. [5]. El-Sayed.A, El-Dahshan Abdel -Badeeh M. Salem, Tamer H.Younis (2009). ―A Hybrid Bolyai, Informatica, Vol. LIV, 2009. [6]. Bainbridge, et al. (2014). Artificial Social Intelligence. Annual Review of Sociology (1994), 20, 407-436. [7]. Woolsey, G and Johnson, J. D. (2011). On Inexpert systems and Natural Intelligence in Military Operations Research. Interfaces- INFORMS (July 1991), 21(4), pp. 2-10. [8]. George F Ludger (2009). ―Artificial Intelligence - Structures and strategies for complex problem solving‖ 5th Edition, Pearson. [9]. Peter Norvig; Stuart Russell, (2016) ―Artificial Intelligence: A Modern Approach". [10]. William Leigh & Allen N Smith, (2016) ―PROLOG to Expert Systems‖, Mitchell Publishing Inc, pp. 1- 3. [11]. L. Thorne McCarty (2007). "Reflections on TAXMAN: An Experiment on Artificial Intelligence and Legal Reasoning‖, Harvard Law Review, Vol. 90, No.5, March-2007.

Research Scholar of OPJS University, Churu, Rajasthan