An Ontology-Based Collaborative Business Intelligence Framework

Muhammad Fahad

and Jérôme Darmont

Univ Lyon, Univ Lyon 2, UR ERIC, 5 avenue Mendès-France, 69676 Bron Cedex, France

Keywords: Collaborative Business Intelligence, Data Visualization, Knowledge Base.

Abstract: Business Intelligence constitutes a set of methodologies and tools aiming at querying, reporting, on-line

analytic processing (OLAP), generating alerts, performing business analytics, etc. When in need to perform

these tasks collectively by different collaborators, we need a Collaborative Business Intelligence (CBI) plat-

form. CBI plays a significant role in targeting a common goal among various companies, but it requires

them to connect, organize and coordinate with each other to share opportunities, respecting their own auton-

omy and heterogeneity. This paper presents a CBI platform that democratizes data by allowing BI users to

easily connect, share and visualize data among collaborators, obtain actionable answers by collaborative

analysis, investigate and make collaborative decisions, and also store the analyses along graphical diagrams

and charts in a collaborative ontology knowledge base. Our CBI platform builds a dashboard to persist col-

laborative analysis, supports interactive interface for tracking collaborative session data and also provides

customizable features to edit, update and build new ones from existing graphs, diagrams and charts. Our

CBI framework supports and assists information sharing, collaborative decision-making and annotation

management beyond the boundaries of individuals and enterprises.

https://orcid.org/0000-0002-7258-9884

https://orcid.org/0000-0003-1491-384X

1 INTRODUCTION

In recent years, there has been a massive and rapid

growth of data. The transformation of large volumes

of data into useful information to help the decision

making process is called Business Intelligence (BI).

According to InfoTech research (2020), BI is

defined as an enterprise-wide capability to capture,

transform and report data or an event into actionable

information to enable fact-based tactical and

strategic decisions. BI plays a vital role in fulfilling

the overwhelming need of large enterprises for

analyzing business data in competitive

environments. Therefore industrialists and

researchers develop BI strategies and tools to enable

reporting and analytics for decision making on large

datasets, strengthen business processes and

operational research activities. BI analytic tools and

technologies help reap the maximum benefits from

business operations and take good data-driven

business decisions. Through well-informed business

decisions with real-time data, organizations compete

with each other and improve business forecasting in

large business clusters.

Typical BI software has features such as

reporting and visualization, trend analysis, customer

behavior analysis, predictive modeling, etc.

However, BI only enables and restricts decision

making features within the boundaries of individual

companies. In addition, traditional BI tools provide

services to individual companies rather than a

network of companies characterized by

organizational, lexical and semantic heterogeneity –

(Golfarelli, 2021). This drawback leads to explore

innovative approaches such as Collaborative

Business Intelligence (CBI) to make collective

decisions incorporating external data beyond the

boundaries of enterprises.

There can be many forms of CBI. It may be a

general discussion among people within companies

or a more report-centric discussion aiming at

commenting and providing feedback on a particular

report (Tackels, 2015). Other forms of CBI may be

seen as adding annotations to specific items in a

report, data visualization, or information sharing,

etc. In addition, through CBI organizations achieve

480

Fahad, M. and Darmont, J.

An Ontology-Based Collaborative Business Intelligence Framework.

DOI: 10.5220/0012131900003541

In Proceedings of the 12th International Conference on Data Science, Technology and Applications (DATA 2023), pages 480-487

ISBN: 978-989-758-664-4; ISSN: 2184-285X

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

information sharing on top of BI tools that include

data warehouses, analytical tools and reporting tools.

CBI enables organizations to gain timely access to

quality information and competitive advantages.

This paper introduces our CBI framework, which

is an accessible, open-source BI platform that

implements the data warehousing process in

Software-as-a-Service (SaaS) mode. CBI users

connect to the platform where they can share and

explore data, help each other to construct data cubes,

and formalize different visualizations via traditional

graphs and charts. CBI users can talk with each

other, build collaborative analysis, comment and

annotate their findings along the compelling graphs

and finally store their analysis in the dashboard. We

also design a CBI ontology that stores annotations,

rather than a database or text files. Ontology-based

storage of annotations indeed makes them machine-

processable, interpretable and they enable high

precision when searching and retrieving knowledge.

The remainder of the paper is organized as

follows. Section 2 presents related work on CBI.

Section 3 details our CBI platform. Finally, Section

4 concludes this paper and hints at future research.

2 RELATED WORKS

There is diverse work in the field of CBI. We

classify the existing literature on CBI into several

categories. The first category consists of research

works focused on collaborative query management.

Giacometti et al. (2011) recommend

multidimensional queries based on a distance

measure calculated by comparing log sessions and

the current session. However, they do not consider

user context or preferences, which are addressed by

Eirinaki et al. (2014). Their system is based on a

lookup mechanism where similar users and queries

help recommend queries to the current user.

Khoussainova et al. (2011) propose another research

work for auto-completion and query management.

They develop a context-aware system that helps

novice users formulating SQL queries. Their system

does not recommend complete queries, but possible

additions to various clauses in the user’s queries.

Sapia (2000) differs from these approaches, as they

use predictive prefetching in on-line analytical

processing (OLAP) to minimize query execution

time. They use a Markov model based on the user’s

multidimensional data. Moreover, Golfarelli et al.

(2012) enhance the decision making process by

sharing knowledge and operational data across

networks of peers. They develop a language for

semantic mappings between multidimensional

schemata of peers, as well as a query reformulation

framework.

The second category we propose consists of

research works that focus on OLAP session analysis

for preference-based recommendations within BI

platforms. It may be difficult to differentiate them

from the first category, but they actually focus on

analyzing previous sessions and not recommend

single OLAP queries. Jerbi et al. (2009) investigate

the interactive and navigational nature of user query

behavior. They extract relevant elements from the

user profile and enrich the query answer before

presenting it to the end-user. Next, Aligon et al.

(2015) present a collaborative filtering approach that

recommends OLAP sessions by analyzing previous

sessions, while not recommending single OLAP

queries. Their work is unique in the sense that they

compare sessions rather than queries. Aufaure et al.

(2013) also use recent analytical sessions to

recommend queries based on a probabilistic user

behavior model and query similarity metrics. They

aim to reduce latency time by the use of a cache

manager that prefetches objects. Wu et al. (2007)

combine the aggregation power of OLAP and

keyword-driven analytical processing. They develop

scalable algorithms for subspace generation, novel

ranking and dynamic facet construction. Cabanac et

al. (2007) develop relational OLAP to add

annotations on multidimensional data. Such

annotations enable decision-makers to share and

communicate data among all collaborators. This

work is very close to ours, but we store annotations

in an ontology rather than a database. Ontology-

based storage of annotations makes them machine-

processable, interpretable and they enable high

precision for searching and retrieving knowledge.

Ontology-based dashboard let different collaborators

to get a better understanding of the data by making it

easier for them to find, manipulate and access

information related to collaborative sessions held

between different collaborators.

3 COLLABORATIVE BUSINESS

INTELLIGENCE PLATFORM

This section details our CBI platform, which is

accessible, open source and free. It implements the

data warehousing process in Software-as-a-Service

mode. In the following subsections, we first present

the architecture of our CBI Framework. Then, we

discuss its implementation and elaborate a use case

An Ontology-Based Collaborative Business Intelligence Framework

481

scenario to demonstrate how our CBI framework

helps collaborators to work together.

3.1 CBI Framework Architecture

The main components of our CBI platform are the

collaborative OLAP, Annotation Management

System (AMS) and User Session Handler (USH).

The AMS and USH components help store

collaborative data semantically into the knowledge

base. By collaborative data, we mean parameters of

data cube (measures, dimensions, filters, etc.),

collaborator personal information and annotations

(comments, opinion, analysis, etc.) by collaborators.

Figure 1 illustrates the architecture of our CBI

platform. With these components, our CBI platform

democratizes data by allowing BI users to easily

connect, share and analyze data, obtain actionable

answers and store their collaborative analyses along

the graphical diagrams and charts in the

collaborative knowledge base.

Figure 1: Our CBI platform.

3.1.1 Collaborative OLAP

The most significant foundation stone in BI is

leveraging OLAP, which permits end-users to

navigate through aggregated data in a

multidimensional data model. It supports various

visualization features, such as creating various

charts, representing tabular data and also standard

operators including drilldown, rollup, and slice and

dice. Different end-users, i.e., data analysts, data

scientists and novice users, can run OLAP as per

their requirements and make a collaborative session

to organize, analyze and visualize data multi-

dimensionally. First, collaborators can connect to the

desired backend database or data warehouse. Then,

data cubes schemas can be designed in collaboration

with teams according to their requirements.

Traditionally, data cubes derive from a relational

fact table that contains some quantitative

metrics, i.e., measures, over which some calculations

can be performed, and dimension tables that are axes

of analysis whose attributes are called members.

Finally, users can visualize cubes to measure and

display dimensions via different types of

visualization diagrams. They can interact via the

USH and collaborate over multi-dimensional data

via the AMS. Both components are elaborated

below.

3.1.2 User Session Handler

Collaborators connect to the CBI platform via the

USH, where they can collaborate with each other.

When a collaborator connects, the USH stores all

user-specific information in the User Profile

Ontology (UPO). It also stores the location and

spatiotemporal information about the collaboration

held between collaborators. The UPO allows the

reusability of online Web ontologies, i.e., FOAF,

(Friend Of A Friend), TimeLine and GeoNames.

Particularly, we reuse a FOAF ontology (Vakaj and

Martiri, 2011) to capture collaborator information in

the CBI platform. The FOAF ontology describes

persons, their activities and their relations to other

people and objects. We also use the TimeLine

ontology (Raimond and Abdallah, 2007) that

captures the temporal information of collaboration.

In addition, we use the GeoNames ontology

(Maltese and Farazi, 2013) to capture either the

physical or virtual location of the collaborative

session. This ontology constitutes a well-known

geospatial dataset providing data and metadata from

around 7 million unique named places collected

from several sources.

3.1.3 Annotation Management System

The collaborative session takes input from

collaborators over the OLAP graphical interface.

Input can be annotations of any type, i.e., question,

answer, text comment or description, and can be of

any form, i.e., general feedback, report centric

discussion, data analysis, task coordination,

information sharing, etc. The AMS takes all the

annotations related to discussion and analysis held

DATA 2023 - 12th International Conference on Data Science, Technology and Applications

482

Figure 2: Top level view of CBIOnt.

between collaborators, and stores them in the

Collaboration Session Ontology (CSO). The AMS

allows several operations such as add, edit, update

and delete, over collaborator annotations.

3.1.4 Cbi Knowledge Base, CBI Ontology

Our aim is to build an efficient and fast storage and

retrieval of information between collaborators.

Therefore, we aim at incorporating ontologies in the

collaborative platform so that different types of

inferences can be achieved on collaborative session

data. We design the CSO to formally describe and

conceptualize the domain knowledge and store

collaborative session data between collaborators.

Data captured in the CSO becomes machine-

interoperable and machine-processable to facilitate

easy knowledge sharing, with common vocabulary

across independent collaborative teams and

organizations. These subontologies UPO and CSO

together constitute a CBI ontology named CBIOnt

(Fahad et al., 2022). For this sake, we use the Web

Ontology Language (OWL), which is the W3C

standard for developing ontologies. Figure 2

represents a two-layered example of CBI knowledge

base. The top layer illustrates OWL concepts that

capture rich and complex knowledge about the

collaboration. The bottom layer constitutes the

Resource Description Format (RDF), which is the

W3C standard for storing Web data. CBIOnt plays a

significant role in improving the effectiveness and

efficiency of the CBI system, which evolves as our

project grows with time.

3.2 Implementation of the CBI

Framework

We use CubeJS

for building our CBI platform.

CubeJS is an open source BI platform that supports

data integration from all major data sources,

designing multidimensional data warehouses and

OLAP navigation. Moreover, CubeJS implements

the GraphQL (Porcello and Banks, 2018) API that

provides a complete and understandable description

of the data. CubeJS constructs data cubes that

exploit a JSON-based metric skeleton to express

data calculations that can be exposed by GraphQL.

There are three tabs in the CBI framework through

which BI users build their collaborative analysis

over OLAP. The following are the details.

3.2.1 Exploration

The first tab of the interface, “Explore”, allows end-

users create various types of visual cube

representations. It allows selecting measures and

displaying dimensions of the data cube. CBI users

can apply filters and choose segments and time

frames to visualize cube data. This tab also allows

users of the collaborative framework to add

comments on the cube via the new “Add Comment”

button. By the new “Enlist Comment” button, CBI

users can see all the comments added during the

collaborative session held among collaborators. One

can add, edit and delete comments from the

cubejs site: https://cube.dev/

An Ontology-Based Collaborative Business Intelligence Framework

483

collaborative session at any time. Once the cube is

formed and comments are added by collaborators,

the “Add to Dashboard” button stores the cube on

the dashboard so that collaborators can use it later

on.

3.2.2 Dashboard

The dashboard enables collaborative analysis

persistence. The “Dashboard” tab helps store and

visualize all the cubes created by collaborators.

Moreover, we enhance the dashboard so that already

created data cubes can be editable at any time. Each

cube is provided with options to edit, delete and

enlist comment buttons. On clicking the edit button,

the cube enters in the “Explore” interface where

updating the cube is possible. When a collaborator

saves a cube along with comments, our platform

persists his/her analysis for future use.

3.2.3 Export Data

With the help of the “Export Data” tab, end-user can

export all stored cubes on the dashboard as JSON

files containing GraphQL queries that allow

reconstructing cubes, particular information (name,

description, etc.) about cubes, and all the comments

added by the collaborators.

3.3 Use Case Scenario

We build our case study upon the Star Schema

Benchmark (SSB; Neil et al., 2009). SSB provides a

data model, i.e., a multidimensional schema

(Figure 3) and a workload model, i.e., a set of

queries as analyses. Thus, we can build graphs and

charts to demonstrate our tool. Let us discuss the

SSB aspects that are necessary for understanding our

use case scenario. The details of generated data and

queries are available from Neil et al. (2009).

3.3.1 Schema and Dataset

In SSB, there are four dimension tables, i.e.,

CUSTOMER, SUPPLIER, PART and DWDATE,

and a fact table named LINEORDER. In data

warehousing, a fact table consists of measurements,

metrics or facts. The fact table may contain many

degenerate dimensions. According to Kimball

(2002), in a data warehouse, a degenerate dimension

is a dimension key in the fact table that does not

have its own dimension table, because all the

interesting attributes have been placed in analytic

dimensions. Degenerate dimensions are essential for

grouping together related fact table’s rows. We only

focus here on attributes from LINEORDER that are

necessary in the upcoming sections.

Measure count calculates the total number of

orders.

Degenerate dimension lo_orderpriority is a fixed

text. Only five values are allowed: URGENT,

HIGH, MEDIUM, NOT SPECIFIED and LOW.

Degenerate dimension lo_shipmode is a fixed

text. Only seven values are allowed: AIR, SHIP,

MAIL, FOB, TRUCK, RIG AIR, and RAIL.

We use degenerate dimensions lo_shipmode and

lo_orderpriority for grouping together related rows

in LINEORDER fact table.

Figure 3: SSB Snowflake Schema.

3.3.2 Use Case

Jean belongs to an organization that casually uses

our CBI platform. She meets Kim at the Data

Summit and has an exchange together. Kim, who is

a novice user, finds interesting to capture knowledge

exchanged within collaborative sessions and benefit

from data visualization and information sharing.

He asks Jean to help him explore some data

(actually SSB’s data). The conversation during the

collaborative session is elaborated below.

Kim looks at the LINEORDER fact table and

inquires what types of mode of shipment are

possible for the delivery of orders, and what types of

order priorities are set by customers.

DATA 2023 - 12th International Conference on Data Science, Technology and Applications

484

Figure 4: Cube for SSB Order.

While looking at the lo_shipmode attribute in

LINEORDER, Jean notices that there are seven

types of shipment of orders, i.e., AIR, SHIP,

TRUCK, RAIL, etc. Immediately, she creates a

cube (Figure 4) that counts the orders and dimension

lo_shipmode to display shipment of orders. Then,

Jean chooses a pie chart representation (Figure 5) to

observe modes of shipments. She realizes that the

most common type of freight transport is TRUCK.

Kim and Jean discuss whether road shipping is the

most cost-effective way to ship orders.

Kim wants to investigate the proportion of

shipment modes. Helped by Jean, Kim adds the

lo_orderpriority dimension in the cube and chooses a

tabular representation of data. Jean tells him that he

can create other types of charts (line graph, bar

chart, etc.) for better visualization.

Kim chooses a bar chart and observes order

priorities (Figure 5). He observes that Urgent and

High demand of deliveries are mostly required when

goods need to be shipped right away and must be

delivered as fast as possible. Both Jean and Kim add

their diagrams onto the dashboard.

Now, Kim is very curious to know whether there

is a correlation between mode of shipment and

delivery priorities, and what mode of shipment is

mostly preferred by suppliers to meet order

deadlines? He restricts his interest modes of

shipment TRUCK and AIR and delivery types

HIGH, LOW and URGENT. Kim comments on the

CBI platform. Immediately, Jean updates the cube,

creates a bar chart measuring the count of

LINEORDER and displays the mode of shipment

and order priorities to observe underlying data. Both

add filters on order priority and order mode of

shipment for some values they want to investigate

(Figure 6).

The conversation goes on between the two friends

until Kim clears his opinion about the mode of

shipment and delivery priorities of orders. All the

conversation between him and Jean happens in the

form of comments that are annotated along the graph

and stored in the dashboard.

Eventually, Kim wants to share his experience

with other team members of his own organization.

Therefore, he asks Jean how he can demonstrate this

to his team. Jean tells him about the “Export data”

feature of the CBI platform, which allows end-users

to export a whole dashboard as a JSON file. He

exports the dashboard and takes it with him. He can

now benefit from the charts, comments and

annotations during meetings with his colleagues.

An Ontology-Based Collaborative Business Intelligence Framework

485

Figure 5: Dashboard containing a pie chart and bar graph.

4 CONCLUSIONS AND

PERSPECTIVES

This paper presents our CBI platform, which enables

collaborative data explorations where BI end-users

easily connect, manipulate data, uncover hidden

facts, make comprehensive overview of data and

present their findings in compelling visualizations.

CBI platform constitutes a dashboard to persist

collaborative analysis, supports interactive interface

for tracking collaborative session data and also

provides customizable features to edit, update and

build new ones from existing diagrams and charts at

any moment. With this feature, dashboards are

available to other collaborators to quickly and easily

see trends and correlations in data anytime and

anywhere, consequently achieving time saving

benefits.

Our CBI platform uses the CBIOnt ontology to

store session knowledge on ontologies as open,

smart, machine-interoperable and machine-

processable data to facilitate easy domain knowledge

sharing, with a common vocabulary across

independent collaborative teams and organizations.

In this way, dashboard connected with CBIOnt is

useful for monitoring, measuring and analysing data

among collaborators, and also enabling efficient and

effective storage and retrieval of session data.

One of our ongoing future directions is to make a

searchable dashboard based on semantic features.

We believe that the semantic layer based on

ontologies shall play a major role in CBI’s research

and development.

ACKNOWLEDGEMENTS

The research depicted in this paper is funded by the

French National Research Agency (ANR),

project ANR-19-CE23-0005 BI4people (Business

Intelligence for the people).

DATA 2023 - 12th International Conference on Data Science, Technology and Applications

486

Figure 6: Order count w.r.t. Order Priorities and Order Shipmode.

REFERENCES

Aligon J., Gallinucci E., Golfarelli M., Marcel P., Rizzi S.

(2015) A collaborative filtering approach for recom.

OLAP sessions, DSS Elsevier vol. 69 pp. 20-30

Aufaure, M.A., Kuchmann-Beauger, N., Marcel, P., Rizzi,

S., and Vanrompay, Y. (2013) Predicting Your Next

OLAP Query Based on Recent Analytical Sessions,

DaWaK, LNCS, vol. 8057, pp. 134–145, Springer

Cabanac, G., Chevalier, M., Ravat, F., Teste, O. (2007) An

annotation management system for multidimensional

databases. DaWaK, LNCS, pp. 89-98

Eirinaki M., Abraham S., Polyzotis N., Shaikh N. (2014)

QueRIE: Collaborative database exploration. IEEE

TKDE, vol. 26 (7), pp. 1778-1790

Fahad M., Darmont J., and Favre C (2022) The Collabora-

tive Business Intelligence Ontology (CBIOnt), BI and

Big Data, RNTI-B-18, pp 61-72

Giacometti A., Marcel P., Negre E., and Soulet A. (2011)

Query recommendations for OLAP discovery driven

analysis, IJDWM, IGI Global, vol. 7(2), pp. 66-90

Golfarelli, M., Mandreoli, F., Penzo, W., Rizzi, S., Turric-

chia, E. (2012) OLAP query reformulation in peer-to-

peer DWH. Info. Sys., vol. 37(5), pp. 393-411

InfoTech (2020), Build a Reporting and Analytics Strate-

gy, https://www.infotech.com/research/ss/build-a-repo

rting-and-analytics-strategy

Jerbi H., Ravat F., Teste O. and Zurfluh G., (2009) Prefer-

ence-Based Recommendations for OLAP Analysis,

DaWaK, LNCS, vol. 5691, pp. 467-478

Kimball, R., Ross, M. (2002). The Data Warehouse

Toolkit: The Complete Guide to Dimensional Model-

ing (Second ed.). ISBN 978-0-471-20024-6.

Khoussainova N., Kwon Y., Balazinska M., and Suciu D.

(2011) SnipSuggest: Context-Aware Autocompletion

for SQL, VLDB Endowment, vol. 4(1), pp. 22–33

Maltese, V. and F. Farazi (2013). A semantic schema for

geonames. In proc of INSPIRE – June 25th 2013.

O'Neil P.E., O'Neil E. J., Chen X., and Revilak S., The

Star Schema Benchmark and Augmented Fact Table

Indexing, TPCTC‘09, LNCS, vol. 5895, pp. 237-252

Porcello E. and Banks A., (2018), Learning GraphQL -

Declarative Data Fetching For Modern Web Apps.

Raimond, Y. and Abdallah, S. (2007). The timeline ontol-

ogy - owl-dl ontology. Technical report,

http://motools.sourceforge.net/timeline/timeline.html.

Sapia C. (2000) PROMISE: Predicting Query behavior to

Enable Predictive Caching Strategies for OLAP sys-

tems, DAWAK 00 UK, pp. 224-233, LNCS

Tackels D. (2015) 6 benefits of a collaborative approach to

analytics & bi, https://www.sigmacomputing.com/

Vakaj, E. and E. Martiri (2011). Foaf-academic ontology:

A vocab. for the academic community. pp. 440–445.

Wu P., Sismanis Y., Reinwald B. (2007) Towards key-

word-driven analytical processing, Proceedings of the

ACM SIGMOD, pp. 617–628. ACM.

An Ontology-Based Collaborative Business Intelligence Framework

487