**2. Temporal knowledge management methods**

To answer the question of how can spatial-temporal knowledge be created in a given business domain, this section explores the aspect of time and space in knowledge creation.

### **2.1 Spatial-temporal knowledge creation process**

In natural sciences, the standard concept of time is determined by what is specified and measured [3]. With respect to the temporal aspects of time, this is measuring time with respect to the temporal duration of an interval or event. Galton in [3] notes that time flows in one direction thus moving forward.

Space is the geographical representation of a real-world object, a place, or a region. Space is said to be moving in any direction.

The creation of knowledge that contains aspects of space and time involves building what is known about spatial-temporal knowledge. This is described as a profoundly interactive process that is affected by other people, the environment, and the actions of objects or individuals. As noted in [4], knowledge is considered inseparable from the temporal process of creation, interaction, interpretation as well as context and space of where creation is completed. It is important to look at how space and time are interconnected within intervals of time or events in a certain geographical space or region.

The creation of space and time-based knowledge has been described in many aspects. The study on how space and time are intertwined within the knowledge creation process is described in [4]. As the researchers noted, knowledge is not information as it requires interpretation and exists between explicit and tactile dimensions. The creation of knowledge involves a combination of various elements of time and the context for which space and time exist. In the creation of knowledge, there are several dimensions to consider (a) object space as to where an object resides, (b) object time as the time when an object is engaged in some event, and (c) the process of deriving the object space and time can be termed as knowledge creation that can then be communicated through various means.

The creation of knowledge can be about relationships as noted in [5], where set-oriented relationships on moving objects can enable the creation of knowledge with respect to properties of movement such as time, speed, direction, accelera¬tion, deceleration, accumulated time intervals as well as the distance within the instance of time.

Others like Yuan [6] highlights events and processes as key to the creation of event-driven knowledge within geospatial research. The researcher discusses the importance of events in scientific inquiries and reviews event-based data modeling and computing. Instead of just applying machine learning (ML) algorithms to readily available data, Yuan highlights the need to add curiosity by asking why events are important and how these can be computed within geospatial research [6]. Events are important to understanding the world and the objects involved in the events.

Based on the domain of service, there are several methods detailed in the literature with respect to the creation of spatial-temporal knowledge. We give an example of a few here.

In social media with the increase in location-based services, massive data sets are generated that contain text, the context of time and geography, that is in social media tweets which can be termed as spatial-temporal messages and contain a vast amount of information (comments, reviews). Users may be interested in knowing information about something they care about but do not want to be bombarded with tweets. To this respect, researchers are looking at methods to get the top-k relevant tweets to a user's interest such as the work in [7] where approximation top k relevant spatial-temporal knowledge can be published and others can subscribe to.

### **2.2 Temporal knowledge representation**

Knowledge representation was initiated as a discipline within artificial intelligence (AI) with a focus on symbolic forms of representation of knowledge so that it can be manipulated and stored on a computer as noted in [3] and most knowledge representation research problems focus on finding a process to encode knowledge so that machines can use it.

There are several factors that are considered in the process of temporal representation. This includes; (a) the ability to include relative and absolute sense of time, (b) relationships between intervals of time, (c) an ability to have context sense of now, and (d) the context of persistence. Allen outlines derived principles for representing temporal knowledge with logic statements that can be used for interference and deduction to enable reasoning about the knowledge represented [8].

Galton highlights several aspects of reasoning that are employed in knowledge representation as rule-based methods using "if… then…" rules which can be used to formulate either data-driven or goal-driven conclusions in the development of expert systems [3].

Model-based reasoning approaches are other forms of knowledge representation that form the explicit encoding of a domain with structural models. Case by case uses existing known solutions for which to build new structures, while hybrid methods combine any of the above representations.

On spatial representation, a landmark paper by Randall et al. [9] details a formal treatment of space and spatial relations within the aspects of time. The researchers introduced the region connection calculus while the work in [10] focused on geographical information systems. A region is an extended potion of space and can be applied at any time. An example using cyberspace, suppose there are two regions R1 and R2, Galton in [3] details relationships that can be used to define the two regions. Now integrating those definitions with time, one can represent the two regions in any event or interval of time. There is some expectation that geographical regions have valid durability within a specific snapshot of the world. As such, it is possible to represent a region in temporal snapshots of time.

With domain-specific encoding of entities of a region, aspects of time can enable representation of critical knowledge that is not obvious. To complete this process requires a comprehensive framework for processing of space and time knowledge and this can be implemented within information systems [3]. Next, we discuss an overview of knowledge processing methods within the spatial-temporal paradigm.

### **2.3 Spatial-temporal knowledge processing**

After creating and representation of spatial-temporal knowledge, next, we examine how knowledge is processed.

Processing knowledge involves the use of information systems. There are several elements to consider in the processing of spatial-temporal knowledge specifically when it comes to the era of big data where exponential volumes of data streams are generated from connected devices and sensors every second from social media, e-commerce, energy, and healthcare among others.

The complexity of processing time-oriented data streams was noted in [11], adding a new dimension of space adds even more complexity in the process of deducing knowledge.

Historically knowledge processing has been characterized within a framework comprised of six phases in data mining literature; business understanding, data collection, data processing, modeling, evaluation, and deployment [12]. Several methods have been suggested in the literature that utilizes this framework particularly in processing spatial-temporal data already created in database systems [6]. However, as noted in [4], data within a database only contain a geometric representation of time and space but not the relationships that are derived by knowledge representation. As such frameworks for processing that data to derive any hidden knowledge are needed.

In recognition that historical knowledge processing frameworks that use [12] do not incorporate methods for processing temporal aspects of data, researchers in [13] designed a temporal knowledge framework by incorporating techniques in temporal abstraction and data mining. This work was later enhanced in [11] to incorporate components for deriving temporal relationships in time-oriented high-frequency data within the data processing and modeling phases. Although the researchers demonstrated their work in clinical case studies in [14], their work did not specify the potential for the inclusion of spatial data and the wealth of extra spatial knowledge that can be added to the knowledge discovery process.

### *2.3.1 Current focus on spatial-temporal knowledge processing*

In most of the research that seeks to process knowledge that incorporates both space and time, the focus has been mostly on data processing and modeling phases. We highlight some of the works here.

Classification: Spatial-temporal access methods for classifying knowledge are detailed in [15] where the researchers note that classification is based on the data indexing. The researchers indicate that there are indexes for historical, current and recent, future as well as the process for which data is processed in parallel or distributed systems.

Clustering: There are various clustering approaches for processing spatialtemporal data. A good review is presented in [16] where researchers defined clustering based on the purpose of use such as hotspots for identifying the high density of some phenomena or cluster trajectories in finding groups of elements

### *A Spatial-Temporal Knowledge Management Framework DOI: http://dx.doi.org/10.5772/intechopen.101797*

that have similar trajectories. An example would be a group of taxis that follow similar routes, or in weather, hurricanes that show similar trajectories, and in the combination of time and space, there are clusters that can be developed for identifying locations that have similar spatial maps.

Rule-based techniques: These are normally termed as methods for temporal associations rooted in sequence mining. In particular, mining relationships in spatial-temporal data aims to identify relations among pairs of regions which may vary by time as noted in [17]. The researchers identify interrelating windows as well as the windows of time within which the interactions happen.

Information flow: In quantifying information flow within a network, then temporal ordering among elements within the network is to be considered. Multiple works have considered the Lagrangian reference frame to study information flow thus quantifying the state of a physical system [18].

Spatiotemporal outlier: This is a technique that has been applied in extracting information that differs significantly from others. As noted in [19], detecting spatial-temporal outliers has many applications, in transportation, ecology, public health, security, and location-based services.

While traditional ML algorithms have been demonstrated as efficient in analyzing and modeling spatially and temporally variable environmental data, automated feature representation and learning in deep learning (DL) models are noted as able to capture spatial, temporal, and spatial-temporal dependencies [20]. These are demonstrated in traffic prediction, where researchers in [21] use deep neural networks to model spatial dependencies and temporal dynamics on large-scale traffic data to predict accurate taxi demand. Neuro networks (NN) are also utilized in [22] where graphical convolutional networks (CNN) are proposed to forecast urban planning and traffic management. Reinforcement learning has been proposed in [23] for spatialtemporal pricing in ride-sharing services by using two feed-forward neural networks. Once spatial models are generated, they are evaluated using some testing dataset and then deployed. The derived special models are the knowledge that has been generated from spatial data and deployment of that knowledge is known as knowledge sharing.

We discuss some of the elements that are involved in spatial-temporal knowledge sharing next.

### **2.4 Temporal knowledge sharing methods**

There are various methods discussed in the literature for knowledge sharing. In this section, we discuss some of the key aspects in spatial-temporal knowledge sharing and highlight limitations still to be addressed.

The competitive advantage of any organization can be greatly enhanced with a well-defined process of knowledge exchange among individuals, teams, and units within the organization. There are several enablers for the seamless flow of knowledge within an organization including secure communication channels that enable dissemination and flow of information to those who require it.

There can be resistance to sharing knowledge among knowledge workers as noted in [24]. These include lack of communication, lack of trust, cultural differences, and personal difference among others.

The creation and processing of spatial-temporal knowledge can generate critical information that would benefit a business, organization, or even a community, therefore, there has to be a well-defined process for which that information is shared.

With the rise of critical questions about how and what are the results generated from models, there needs to be a process to freely share the knowledge creation process including what data was used and how it was acquired, what are the features included in models as well as explaining the results of the models.

Sharing spatial-temporal knowledge also raises the bar on how trustworthy that knowledge can be taken, interpreted, and applied within any domain.

With sharing more information about the spatial-temporal knowledge creation process, this can contribute to user buying when the results of models are to be utilized to make a critical decision about people, places, and communities. For this to be accomplished requires well-structured frameworks for knowledge creation, processing, and sharing. We present some of the existing frameworks suggested from the literature in the next section.
