The advance of artificial intelligence (IA) has revolutionized many sectors and tourism is no exception. The growing digitalisation of the sector has allowed advanced IA technologies to improve user experience, optimize service management and facilitate strategic decision-making. That's how we have a system. SYSTEM-TH-001 an innovative solution that integrates automatic learning, knowledge management systems and smart agents to improve customization with efficiency in the tourism industry.
Automatic Learning (ML) is one of the key technologies in this area. Inspired by human learning, ML allows the detection of patterns in large data volumes and provides precise predictions or recommendations. Qualification models such as Vectorial Support Machines (SVM) and neural networks have been shown to be effective in customization of offers with customer segmentation for the tourism sector.
In addition, knowledge management systems (KMS) play an important role in organizing and effectively recovering information. Those systems allow for high data production and easier access to useful information about decision-making.
On the other hand, smart agents designed under methodology SOFTWARE-TH-001 they allow dynamic interaction with and between the environment, offering automated solutions that have been adapted to SYSTEM-TH-001. That bioinspired approach facilitates job optimization and efficient workload distribution.
Overall, these technologies turn to SYSTEM-TH-001 in a robust and scalable system that redefines how tourism can benefit from artificial intelligence, improving user experience with best management of available resources.
2.1 Introduction
The artificial intelligence technologies used in the system are explored with this chapter SYSTEM-TH-001 and with an emphasis on the state of art in tourism, as far up to date. Automatic learning is addressed by comparing its similarity with human training and their application to identify patterns in large data volumes, improving decision-making. Classification models such as SVM and neural networks are discussed in detail. In addition, information and recommendation management systems have been discussed with a view to optimizing tourism experience. Smart agents, their ability to interplay with the environment and its implementation are also addressed SYSTEM-TH-001 using the methodology SOFTWARE-TH-001.
The state of art will be analysed in each area, providing a theoretical and practical framework that'll allow us to understand how important artificial intelligence can have an impact on digital tourism's evolution.
2.2 Automatic Learning
Human and autolearning, while they seem different have a set of similar principles or processes. Basically they require information acquisition and interioring that leads to skills development, knowledge with the capacity of adapting them for new Langley situations ( 2022 ) Botvinick et al. ( 2017 ).
From a philosophical perspective, human learning goes beyond mere skills acquisition and involves character formation with ethical development while critical thinking capacity. Aristoteles, in his work: "Etima de Nicomac '(The Ethics of Nicamaco) looks at the importance and practice with respect to developing virtues by concluding that Learning is an ongoing process for improving personal development. Wang, Guo and Chen ( 2022 )
Automatic learning, ML could understand as an extension currently complementary to human Learning. At ML machines learn, acquiring knowledge and some skills through software we call them algorithms to simplify it that allow the machine incremental learning throughout time by means of practice. ML is the area that, using data and mathematical models as a means to identify patterns for improving decision-making by Meltzoff et al. ( 2009 ).
That parallelism isn't a recent dubbing, and in fact one of the pioneers at an intertidal intelligence level (Alan Toring) has already recognized their potential.
programmable machines to learn from experience and improve performance at expectations tasks Naha or Murphy ( 2015 ). If we deepen more knowledge of artificial intelligence, We can observe great efforts to develop learning systems that emulate or even exceed the cognitive skills from Lake et al human Learning. ( 2016 ) Romero-Mendez et al. ( 2023 ).
One of the most appreciated ML characteristics today, is his ability to analyze large amounts and data for how we look at patterns that are far from what humans Russell or Norvig have been enjoying ( 2010 ).
That makes ML a very significant advance in the area of artificial intelligence, as it allows machines to learn how they define them and better adapt with changes. That capacity offers the potential to build innovative and efficient solutions throughout a number of areas, including tourism. Using ML the tourism sector can benefit from improved data analysis processes, recognition of patterns and custom recommendations with automatic task processing that may lead to better customer experience as well how best they have been achieved through an increased performance by Afsahhosseini's or Al-Mulla ( 2020 )
The ML area includes three types of learning approaches: monitored, unsupervised and supported. In supervised learning, the model is trained with a dataset where inputs are linked to their relative expected output values and enabling models of how they can learn existing patterns or relationships in data. That knowledge will accordingly be applied for predictions about new Liu-Wu ( 2012 ). The unmonitored system focuses on the discovery of patterns and relationships in a dataset that does not have been tagged, namely without guiding Hinton or Sejnowsky's predefined output values ( 1999 ). Strengthened learning allows models to interacted with the environment, receiving feedback as rewards or penalizes and tipping their actions so that they can maximize an expected return from François-Lavet at. ( 2018 )
The success of an ML model is measured by its ability to effectively solve a concrete task, as can be classification and prediction. Tasks that have high added value in all industries, as can be tourism with the use of data analysis technologies and classification. 2020 ) Afsahhosseini and Al-Mulla ( 2020 ).
The conduct of data analysis using ML allows to discover complex hidden patterns and relationships in the dataset, thus facilitating information decision-making.
Mada. For example, by recognition of patterns and classification techniques they allow them to identify many variables from a journey's experience as those that come with customer preferences or how we buy but also the characteristics for Shafiee & Varjani contracted services ( 2016 ). Projection techniques use historic empresas data to predict future behaviour, enabling the company's better adaptation for change by responding with changes in demand and prices among many other Norbert factors ( 2011 ). Also with ML you can automate tasks and processes from the company in an intelligent way. Also, the recommendation systems that use ML algorithms and provide best recommendations leading to an improvement in market's experience with their satisfaction. ( 2014 ).
Although ML's potential in the tourism sector is significant, it remains vital to use them and consider some ethical implications as well with respect for sustainable development of this technology. Attention should be paid to data privacy, algorithmic biases and the social impact of these systems so that benefits from this technology are available in a sustainable manner beneficial for all decision-makers Vesni ″ c-Ultievi ́ c et al. ( 2020 ).
2.2.1 Classification models
The classification models are a fundamental category of algorithms in the area and statistics. The main aim of this information is to assign a label or category for an observation, based on some predictive characteristics and variables. These models are widely used in a number of areas, such as medicine and financial care. They have been put up with information about their ability to accurately predict or categorize them at an earlier stage)
Main models of classification
There are multiple models of classification, each with its benefits and disadvantages depending on how bad the problem is or what data's available. Some of the most common include:
Logistics regression
Logistics regression is one of the most simple and effective models for binary classification problems. It calculates the probability that an observation will belong to a given class using its signmoid function. It's particularly useful when the relations between predicting and target variables are linear.
Vectorial Support Machines (SVM)
The SVM are powerful models that seek an optimal hyperplane to separate classes from feature space. They use "kernel 'to manage non-linear data, making them versatile and precise in many contexts.
Support Vector Machines, SVM have been re-established as a powerful and versatil machine learning algorithm for an varied set of applications from image recognition to natural language processing at financial prediction with bioinformatics. They have been put up there for the years 90 and as a result of its implementation with success, due to the great variety from problems that it can analyse even more effectively than other technologies. The main advantage of SVM is its ability to find an optimal hyperplane separators that maximizes the range between classes, generating a solution involving only few support vectors and his property which will simplify their application in long-scale problems Mocerza y Muñoz ( 2006 ).
In the financial domain, SVM has shown its effectiveness in detecting credit card fraud. Compared with other classimators, SVM can solve both binary problems of linear and non-liner classification by identifying the hyperplane that best separates input data C Li et al. ( 2021 ). For example, the use of SVM in credit risk techniques has shown a high capacity for adaptation and an efficient global solution to decision-making with this S.-L. problem Li et al. ( 2006 ). Other challenges with its success have been solvency problems or even studies about Nazareth and Reddy cryptomonedas ( 2023 ). The ability of SVM to manage data with high dimensionality and noise puts this technique as a useful tool in the financial environment, but its success has also come into other areas including recognition for images or problems about citizen security among many others that show their great versatility Mocorro y Muñoz ( 2006 ).
SVM has been shown to be a highly effective technique, but the proper selection of its kernel function and an appropriate select with hyperparameters are crucial for optimizing their performance as stated by Bing-yu et al. ( 2009 ). Specifically within the project SYSTEM-TH-001, this method has been investigated as a tool for the detection of gender from text and its functionality but that can be key to his future. The SVM research area remains fully advanced and we can expect new advances with applications leading their boundaries further.
Decision Trees
Decision trees are intuitive models that classify observations through a hierarchical decision structure. They are easy to interpret and useful for problems with categorical or numerical data. When they are combined into random forests (Random Forest), their accuracy and Robustness greatly improve.
Neuronal networks
Neural networks are models inspired by the human brain, ideal for complex and non-linear problems. They use multiple layers of neuroons to learn hierarchical representations from data, making them highly effective in image classification and natural language processing.
Evaluation of Models
The efficacy of the classification models are evaluated by using metrics such as accuracy, recall and F1-score with an area under REC (AUC-ROC). Moreover, it was crucial to avoid problems such as overlays by using technologies like cross-validation and regularization.
2.2.2 Optimization models
Classification and optimization models are a fundamental category of algorithms with regard to automatic learning as well at statistical level. The main goal of this information is to assign a label or category for an observation, based on some predictive characteristics and variables while at the same time they seek best model parameters with optimum performance. These models are widely used in a number of areas, such as medicine and financial care. They have been put up with information about their ability to accurately predict or categorize them at an earlier stage) There are multiple models using optimization technologies to improve their performance. Some of the most common include Vectorial Support Machines (SVM), which are powerful models that seek an optimal hyperplane to separate classes from characteristic space. SVM optimization relies on maximizing the range between classes, guaranteeing better generalisation. To manage non-linear data, SVM use "kernel 'such as Gaussian or Polynomic kerel that transform the datasets into more dimensional spaces where classes are inseparable. Moreover, logistics regression is one of the most simple and effective models for binary classification problems. Optimization is achieved by minimizing a cost function with cross-entropy, using methods such as Stocastic Gradient (GDS) decline. That approach ensures that model parameters converge to an overall optimum.
The decision trees are hierarchical models that classify observations through a number of decisions based on specific characteristics. The optimization in these models involves selecting divisions that minimize metrics such as entropy or Gini Index. In random forests, performance is optimized by combining multiple trees with bagging techniques and reducing the risk of overtuning. Neural networks are models inspired by the human brain, ideal for complex and non-linear problems. They use optimization techniques such as gradient decline, fringes with moment and advanced algorithms like Adam. These techniques suit the weights of connections between neurons to minimize loss function and improve performance. Also models such as Gradient Boosting, AdaBooth and XGBOost use iterative optimization techniques to adjust simple model forms including decision-making trees in sequence. Each iteration fixes errors from the previous model, minimizing an cumulative loss function.
Support optimization Vector Machines
SVM are algorithms with performance based on the use of kernel. A K (x, and) kernel is a real value function N: X × X→ R that acts as an incremental product in actual vector space Z. To this end, there is a role:
0: X → Z, ( 2.1 )
so that K (x, and) = 0 [g] TΦ {y}. X and Z are respectively known as entrance spaces with characteristics.
SVM belongs to the type of algorithms based on Mocorro et al regularization theory ( 2006 ). These methods allow the construction of classification functions to solve an optimization problem in Tikhonov and Arsenwith ( 1977 ):
N X
m ́ ın f 1 g
i = 1 L (yi, f) and g 2 K ( 2.2 )
where μ > 0; HK denotes Hilbert's playback kernel space (RKHS) associated with the k-kerel, and pdg as a rule of f at Rekys. 1, + 1 } indicates the two possible classes of xi and finally L (yi, f) is an error function. In this context, f (x) = 0 It's a decision surface at HK. The typical SVM approach uses the L-specific error function called hinge loss, as defined by:
L (yi, f) = 1 − yi f (xi) and 2.3 )
with (x) and = m ″ ax 0 ).
The problem ( 2.2 ), μ helps to put in place a commitment between an adjustment of the f-solution with data as quantified by L and how complex that's function will be.
It's right to prove that 2 K = ‡ w 2, where w = pn i αiΦ (xi) and 0 is the Mapping as defined in 2.1 ). Also, the problem ( 2.2 ) can be reformulated as:
N X
m ́ ın w, b 1 g
i = 1 ( 1 − yi (wTΦ) + b and 2. ( 2.4 )
The problem ( 2.4 ) is equivalent to solving the following problem of dual optimization Mocerza et al. ( 2006 ):
g ″ n.n β 1 2 β TQß − eTS
( 2.5 )
s.t. and Tβ = 0,
0 ≤ β < C,
where β = (β1,..g) T and 0 It's a vector of zeros, and it is an
some, Q is defined as N × g a positive symmetric matrix with Qi j = yiy JK (xi and x). 1 2 μn's a constant. That problem was convex and quadratic, with every local minimum as well a global level.
For the sake of simplicity, g (β) is defined as a gradient for an objective function with this problem. 2.5 ), that's g (β tema = Qß − e.
A β vector is a stationary point of ( 2.5 ) whether and only if there is a number d with two non-negative vectors λ or μ so that:
g (β) + dy = λ − μ,
λiβi = 0, i = 1..n,
( 2.6 )
μi (C − β) = 0, i = 1..n,
λi, μI ≥ 0, i = 1 No.
Can be shown, Chen et al. ( 2006 ), that a β vector as the 0 ≤ β i < C meets the conditions ( 2.6 ) If and only if
− yig (β) i ≤ d,
− yig (β) i ≥ d, 2.7 )
where Iup (β) ∞ {t ю 124; βg < C, and lt = 1 or β t > 0, and t = − 1 } and
Ilow (β) 1 or β t > 0, and 1 }. ( 2.8 )
Of ( 2.7 ), that m (β] ≤ M {α} is argued where:
m (β)
M (β) ∞ m ″ ın i ю Ilow and {α} yig, 2.9 )
In SVM literature, the problem ( 2.5 ) is resolved using so-called Minimum Sequential Optimization (SMO and Joachims, 1999; Platt, 1998 ). These algorithms are essentially Newton-type quadratic methods that, in order to make the problem computationally treatable have been considered only a subset of variables at each iteration and so called work ensemble rather than working with an entire Qmatrix.
Theorema 1. Let {β k} be an infinite sequence generated by being the infinite sequence created with a BMO-type method for problem ( 2.5 ). Then if Q's a positive defined matrix, the {β k} limit point is an overall and unique minimum problem ( 2.5 ).
Proof. See Chen et al. ( 2006 ). □
The following corollary is accordingly maintained.
Corollary 1.1. If {β k} is an infinite sequence, then the following two limits exist and are equal: l ″ ım K → ∞ ma (βks) = L ́ μ.g. 2.10 )
Taking into account the definition of Iup and Inlow sets from an equation ( 2.8 ), and particularly of β k involved within each set (inside the boundaries] for g big enough with a small tolerance ″ > 0 The following condition shall be met: m (β k) − M [σg] ≤ ∞, 2.11 )
That's:
L ″ ım k →
m (βk) − M [β ks] = 0.
Based on these results, SMO-type algorithms implement their
Retirement criterion:
m (β k) − M [βg] ≤ ∞. 2.12 )
2.3 Natural Language Processing
Natural Language processing, PLNs is an artificial intelligence subarea that focuses on computer and human interplay through language. It allows machines to understand, interpret and produce natural language. The development of this field is being dynamic and evolving very quickly, generating quite significant advances in recent years with applications at different domains such as translation from languages to analysis about feelings or interface focused on Agarwal ( 2019 ).
PLNs develops an extensive set of technologies and approaches, including understanding with a generation from the natural language as well how to conduct speech. Those techniques often use the capacity of MOIs and deep learning as well with other computer technologies for human language analysis, understanding or generation. Currently PLNs shows very significant advances, one of them is the ability to understand structures and meanings for datasets in natural language. Techniques such as analysis or parting) labelling but semantic analytics are how a computer can conduct itself almost about human from an Young et al perspective ( 2017 ).
Deep Learning (DLs) has revolutionized the field of PLNs, enabling models to be created that learn about complex and large-scale information representations. Many times neural networks have a performance more than other methods, tasks such as translation or text generation are very efficient with this type of Lan et al models. ( 2024 ).
Recent advances at PLNs include introduction of LLM, large language models as SOFTWARE-TH-003 and GPT, they have shown an impressive capacity in a multititle of Yenduri et al. ( 2023 ). These pre-trained models can be adjusted by means of a fine-tunnning process for specific tasks, effectively increasing the rich knowledge about language acquired during pre- entry.
PLNs, continues to evolve and researchers or users are currently facing challenges such as improving models with robust performance.
training with other fields, such as multimodal learning and representation of Khurana et al. ( 2022 ).
2.3.1 Identification of gender in texts
The gender classification of a text has advanced research in the area. One of the most common applications is to detect a person's gender that has written an Mukherjee and Balat text ( 2017 ); Huang et al. ( 2014 ), e-mail Corney et al. ( 2002 ), Kucukyilmaz et al. ( 2006 ) posts at Yan and Ja's websites ( 2006 ), contributors to online collaborative projects Vasilescu et al. ( 2014 ); B. Lin and Seebrenik ( 2016 ); Terrell et al. ( 2017 ); M. Das et al. 2019 ) or to detect the gender of users' feedback on CRM Lau and al platforms. ( 2005 ); Amado et al. ( 2018 ). The production of authors' texts has been performed under several Kocher and Savoy languages ( 2017 ); Fatima et al. ( 2017 ); Lotipez- Santillan et al. ( 2020 ), which confirms the benefits of methods that are content-based. Other approaches to solving this problem could be using Kretschmer and Needle graph analysis ( 2005 ); Rangel and Rosso ( 2016 ) as well and the labelling of Part Of Speech (POS. 2019 ).
At Krüger and Hermann ( 2019 ), the authors have been 215 previous jobs published between 2017 and 2019 about the identification of authors' gender. According to this review, the best experiment in previous research literature indicates an accuracy of 93.4 % of the cases extracted from SERVICE-005 Markov et al. ( 2017 ). It's worth mentioning that all research publications considered by this survey conduct standard machine learning procedures to prepare text entry, including topenization and standardization of the Eisengstein ( 2019 ) (such as lemization or stemming. Just as we explain in a Section 5.3 with this approach, we have introduced changes to the standard procedures for improving performance in gender identification as suggested by other previous research Uysal and Gunat ( 2014 ). In addition to this, one of the common limitations for other studies is that only considering binari indentification (will deepen these ideas in Section 3) 5.4.4 ).
According to Krüger and Hermann ( 2019 ), there's a high variability in the accuracy of gender identification with different languages. A deeper look at previous jobs in this review reveals that experiments with languages providing a rich information about the gener, such as Spanish and Portuguese have better performance than English. The new method, useful to SYSTEM-TH-001 we seek to confirm if it's possible with some languages specific characteristics for the purpose of performing gender identification.
Identification of a text's gender, that is to decide about an individual.
described in a text, have been given comparative care with previous studies. Nevertheless, there are applications for targeted advertising Jansen et al. ( 2013 ), evaluating gender differences in online jobs markets Fung et al. ( 2018 ) and recognition of designated entities (NER. 2013 ). In contrast with previous approaches, predominated by Wais database services ( 2016 ); Santamaria and Mihaljevi ́ c ( 2018 ), most recent by S. Das and paik ( 2021 ), show the usefulness of applying machine learning algorithms that analyse contextual information. The proposal for a SYSTEM-TH-001 That's to combine machine learning algorithms with retention of morphological elements that contain information about gender in some languages, for an improved tagging performance from a person described by text.
2.4 Knowledge management systems
Knowledge Management Systems, Knowledge management systems (KMMs) is an area with a specialized approach and often integrated into other more comprehensive knowledge-based system for managing Grace & Butler's information resources. 2005 ). These systems are designed to store, process and use the knowledge necessary for enabling Thampi ( 2008 ). More specifically, KMMs are typically the haven for an organisation's knowledge by using repositories such as databases and data lakes or other structures but they also allow us to analyse this information that has been stored in search of trends while supporting its decision-making from a strategic perspective Sensuse et al. ( 2018 ).
One of the key functions for KMMs is to increase an organisation's effectiveness and efficiency by overcoming challenges that arise from providing information about its relevance, a process achieved with partir system skills in order To generate multiple levels at best as soon after new knowledge acquired Thampi has been integrated ( 2008 ). KMics has multiple applications across several domains, including document and content management systems with e-learning platforms as well a communication network or collaborative work system Grace's ( 2005 ).
For example, document and content managers are focused on providing users with real-time access to the Organization's stored contents typically organized by themes. These systems are implemented with metadata structures that facilitate recovery and navigation between Sensuse et al. ( 2018 ). The E-learning systems are focused on educating and training users by delivering structured, organized knowledge
A friendly way. These systems are designed to improve the learning process by encouraging user commitment, enabling new skills and capabilities gain their best share of development as a whole with Grace's human capital ( 2005 ) Malingkas and Ce ( 2020 ). KMMs have also been re-established as a very popular approach for organisations that seek to promote collaboration, socialization and integration of their human team. K. Lee and Hawamdeh ( 2002 ). These systems can be found in a number of application domains, among them we have been using Thampi ( 2008 ), educational games Grace and Butler ( 2005 ) and increased reality applications C K. Lee and Hawamdeh ( 2002 ). They are also integrated as communication networks, and they serve to share with them Iiriemiokhale's knowledge ( 2020 )
Nevertheless, a successful implementation of KMMs and their integration with or even sharing internal learning processes can be complex as it requires considering multiple factors critical to success including system alignment for strategic objectives but an active interplay about employees. K. Lee and Hawamdeh ( 2002 ).
One example of KMMs is framework ' SYSTEM-TH-001 That's also a documentary and content manager that organizes information related to several activities in the domain of tourism with an aim at facilitating knowledge detection as well.
The exquisite implementation of an KMMs requires a profound knowledge and understanding about social or economic factors that can influence sharing information with one organisation, as well how strategic alignment will be achieved to the objectives for learning from them while managing their familiarity. K. Lee and Hawamdeh ( 2002 ). Also, with an emphasis on the importance of individual level understanding and motivations after knowledge-sharing actions currently occurring studies have identified several factors that can impact this technology's success including user characteristics such as its direction to share information or their confidence in his system but also power dynamics. also about the nature of shared knowledge and how Njiraine ( 2019 ). To ensure an effective implementation of KMMs, organisations have to carefully consider these critical factors and develop strategies that increase the culture as a means for enabling them. They align their system with its strategic objectives while interacting at users by using activity about how they are interested in businesses H. Lin ( 2007 ).
2.4.1 Recommendation systems
Recommendation systems are designed to provide custom suggestions for users by creating a unique profile with each individual, based on an array of different parameters that reflect their interests and preferences. These user profiles have been crucial to enabling recommendation systems To adapt and meet the dynamic nature of users' needs or preferences. Peng and Xu ( 2022 )
The process of building user profiles typically involves data collection from several sources, such as registration information and browsing history. 2016 ). That dataset is analysed to identify patterns that will make it possible for users Hung and Huynh ( 2019 ). Since user preferences can change over time, profile construction allows better adaptation to that shift by generating best recommendations and creating a more advanced use experience of Zhuhadar users with Nasraoui ( 2010 ).
Recommendation systems can be put into place to use different recommendation methods, for example an approach with a good reputation or relevant content recovery paudel et al. ( 2016 ). By considering these different factors, they can provide users with a diverse set of recommendations that cover their interests and preferences. Peng ( 2022 )
The development of an adaptive and precise user profile is thus a critical component in the effective design for recommendation systems as it allows these system to adapt with them, while providing custom suggestions aligned by Kanoje et al users' preferences. ( 2015 )
For these reasons, the ability to create and maintain a set of user profiles over time is one key for final success with an endpoint system because that's why it allows providing efficient suggestions by improving users satisfaction paudel et al. ( 2016 ).
For the framework SYSTEM-TH-001 It's a recommendation system that can make recommendations to the travel agent about customers with more propensity for purchase, and so it may be classified as an relevance-based approach.
2.5 Smart agents
Smart agents are software entities that can interacted with their environment and other actors within them, exchanging information or showing conduct simulates the complexity of real world S.-M. Lee and pritchett ( 2008 ). Intelligent agents, who have autonomy and capacity to interacted with their environment can effectively conduct organisational tasks for achieving Itaiwi et al. ( 2012 ).
The dynamic nature of agent-based systems, where agents assign them with multiple roles and responsibilities can simulate complex scenarios that are trying to model the world. By modeling agents based on the characteristics of today's society, researchers can study conduct that appears with interplay between actors and this offers a valuable knowledge about properties from complex adaptation systems Ahmed et al. ( 2005 )
Smart agents currently designed as physical robots or chatbots are a promising alternative to high-level human assistants, since they have the capabilities of autonomous navigation through complex and dynamic environments that use their knowledge with his reasoning skills for achieving an Zset. Peng ( 2021 ). These agents are proactive, independent and independently based on their own knowledge of predefined rules to solve problems Burgin or Dodig-Crnkovi ́ c ( 2009 ). Your interplay capacity can be used to solve complex problems or distribute workloads with a degree of Burgin and Dodiga-Crnkovi ́ c ( 2009 ). The lifespan of intelligent agents includes meeting a set and targets through available knowledge with an association ( 2.1 ) Burgin and Dodiga-Crnkovi ́ c ( 2009 ). These goals may be independent or have been organized with subocjectivates that allow them to complete high-level targets when they are achieved. The objectives are linked to tasks performed by the agents, both goals and jobs have been elements of an agent's mental status. That serves as a means for storing rules or knowledge that allow performing challenges with Burgin & Dodig-Crnkovi ″ c targets ( 2009 ).
The organization in the form of Multi Agent MASs Systems is open to collaboration, competition and negotiation by using standards for modelling technologies with Zagent-based software engineering. Peng ( 2021 ); Francis et al. 2023 ); Halhenborg ( 2009 ).
The agent-based modeling, Agent-Based Modelling (ABM) Uddin et al. 2021 ) and agent-oriented software engineering, Agens - Oriented Software Engineering
Figure 2.1: Main concepts of SOFTWARE-TH-001 for the development of MASs.
ring (AOSE) Garro et al. 2018 ), are standards in the domain that elements and entities provide their implementation. Some well-known approaches using MASs to solve complex problems or simulate real environments are, the drills of road traffic Fernandez- Isabel et al. ( 2020 ), the distributed systems of support to decision Migulez-Fernández et al. ( 2020 ), bio-inspired machine training systems heads et al. ( 2020 ) and video games J. Wang, Hong and al. ( 2022 ).
MASs are usually designed using specific artifacts and pa-
To conduct the different phases of complex system development. That allows to generate relationships and interactions between agents that are represented graphically, then automatically transformed into source code. SOFTWARE-TH-001 GAIA, Probetheus and Troils are best known agent methodologies at Siswahyudi et al. ( 2018 ).
The agent platforms, source code libraries of some programming languages are typically the solution for applying MAS that have been previously modeled. Highlights in this area are SOFTWARE-TH-004 (Java) Bergenti et al. 2020 and MESA (Python) Kazil et al. 2020 ).
For the development of SYSTEM-TH-001 The agent methodology used has been INGE- NIES, adapting the instrument model to MESA framework for its implementation. The agent model is a bio- inspired and distributed models based on an ant colony's organisation, to generate MAS with multiple ants that work together or independently as per system workload.