A Context-aware Strategy for Mobile Services in M-commerce

International Journal of Database Management Systems (IJDMS) Vol.12, No.3, June 2020
DOI: 10.5121/ijdms.2020.12303 29
A CONTEXT-AWARE STRATEGY FOR MOBILE
SERVICES IN M-COMMERCE
Kuen-Liang Sue
Department of Information Management, National Central University Taoyuan, Taiwan
ABSTRACT
Due to mobile communication technologies, the novel mobile commerce is coming true. The context aware
service is one of the most potential mobile commerce. The data about mobile users and their surroundings
are recognized and collected, which are used to provide the services they may be interested. Fuzzy logical
processing model are utilized to make basic data correspond to the mobile users’ high-level information,
so the recognition of their behaviours and situations could be inferred in our research. The numerical
examples are also illustrated in our scheme. Based on the results of our activity recognition mechanism, it
is possible to automatically discover and transmit the contents or services that users require in specific
scenarios. Therefore, more and more novel mobile applications are coming true. To sum up, the proposed
scheme is supposed to enhance context-aware services and provide more attractive life styles.
KEYWORDS
Mobile commerce, Context-aware service, Activity recognition
1. INTRODUCTION
Fourth Generation (4G) and other mobile technologies are well-developed and smart phones
become more and more available to all. Therefore, people could have a modern living via the
development of mobile communication networks. The mobile services based on the cellular
networks have become a novel tendency, so mobile devices with services are becoming essential
for smart life styles. Furthermore, more and more attractive mobile services are developed and
make a new commercial style, which are call mobile commerce (i.e. M commerce).
Nowadays, mobile users could widely receive mobile applications which is becoming a popular
business and creating the opportunities for mobile commerce due to the widespread of mobile
computing. The context-aware service is one of the important mobile applications. Context
aware applications are provided to the user by discovering and utilizing contextual information
that characterizes the user’s situation. Hence, if the context-aware application gets well
developed, users’ real-time demands or wish could be satisfied immediately and automatically.
For example, when people are shopping in supermarket or department store, they can
automatically get the digital advertisement or promotion information.
The context aware attributes and its framework should be well-organized to store, manage and
utilize the context-aware information. Therefore, this research aims to estimate the mobile user’s
scenario by using basic sensor’s data. To closely meet the requirement in real world, the
proposed scheme could be adaptable by setting proper input or output items and has good
flexibility. Most importantly, the concept of context-aware services could be realized through the
novel strategy in the paper.

30
2. RELATED WORKS
The location-based service is a well-known mobile service, which is one type of context-aware
services. In brief, the context-aware service means a smart service providing according to the
information about the mobile user’s surroundings collected widely and automatically. The
suggested services are judged to be attractive to the user under such scenario [1]. Hence, how to
design the framework to support the smart promotion would be a key job. Moreover, to realize
smart computing anytime and anywhere, a novel information technology called ubiquitous
computing become hot issue in network fields[2]. This is an important field to support the
oncoming context-aware service and mobile commerce [3].
Korpipaa et al. designed the context-aware attributes with the organized structure to manage the
context-aware information systematically [4]. In addition, they classified the factors in an
environment into sound, light, temperature and humidity and measured these value. Thus it is
possible to obtain the high-level information from those basic factors. Whenever a user wants to
query high-level contents or service providers would like to push services to the user, the context
manager would reply to them according to the reasoning situation based on the sensing values of
sensors. The progress for implementing such smart scenario is still an important issue on going.
Moreover, Mühlenbrock et al. tried to develop applications for the support of workplace
interactions [5]. They utilized laptop, handset, personal digital assistant, ambient sound, and user
feedback to collect basic sensor data. The activity, availability and inputs are also defined in
their contribution. Furthermore, Azkune and Almeida proposed an unsupervised recognition
technique [6]. Their study enabled users to label anonymous activity events by showing
contextual information gained from the sensors in house.
Due to the literature survey concluded above, a smart approach to estimate high-level contents
via the sensing value of sensors will definitely become the key point of developing the context
aware service. To achieve goal, the raw data measured by surrounding sensor must be used to
infer the user’s need a concluding and reasoning mechanism. It needs to combine various factors
related with the surroundings. In other words, an inferential strategy is required to estimate a
user’s uncertain situation. The estimation could be solved by means of fuzzy logic. Therefore, it
is feasible to take advantage of fuzzy logic for reasoning context-awareness. The well-designed
model with fuzzy sets will be able to match the requirement to estimate human behaviour in real-
world [7].
Based on spatial reasoning decision support, the fuzzy logic has also been used for location
tracking [8] [9]. Wu et al. analysed the multiple attributes of the product with respect to the
consumer’s desire by a fuzzy approach and classified these products into different levels of
preference [10]. The proposed strategy in our investigation uses fuzzy logic to compute and get
flexible high-level contents from raw data. The activity recognition means to guess or make sure
what activity people are engaging in. Based on fuzzy logic approach a flexible strategy for
achieving the activity recognition is presented in this paper.
3. THE FRAMEWORK FOR CONTEXT-WARE SERVICE
This study is constructed based on a context-aware framework. A mobile station (MS) which is
usually a handset or a pad with a mobile user must be served under a base station in some cell
and could roam to different cells. A base station controller (BSC) will coordinate different BSs to
support mobility of MS. Moreover, a mobile switching centre (MSC) manages different BSC and
mobile databases to provide roaming in a mobile network [11]. To provide context-aware

31
services, a context server is collaborated with the BSC. To enables a MS know its own location,
the location information about the user can estimated from MS’s located cell or collected from
Global Positioning System (GPS) as necessary.
Figure 1. Framework of context-aware service.
Various data of context types are assembled by the context server. Moreover, the values of these
data are converted to the corresponding membership degrees. Accordingly, activity recognition is
preceded in the next step and the context server is engaged to provide suitable services in the
estimated scenario. The proposed context-aware framework in the study is shown as Figure 1. It
is necessary to define the input attributes for users’ activity recognition in a mobile environment,
including background noise, level of positioning signal, moving speed, working circumstance,
and usage frequency of device. The scenario of a user will be reasoned based on these attributes.
Furthermore, if a user feedback mechanism is also considered, it is possible to adjust and rebuild
the membership degrees as well as the concrete rule base. By the way, a more and more accurate
model through continuous training and adjustment can be constructed on the context server.
To sum up, a novel mechanism of recognizing context-aware information from the basic sensors
data is proposed in this study. From the viewpoint of the mobile service provider, it becomes
feasible to recognize a user’s scenario and provide appropriate context-aware applications
through the proposed mechanism.
4. THE PROPOSED SCHEME
The processes in the proposed scheme contain basic context types, membership functions and
fuzzy sets. It also involves the quantitative method of background values in the real world and
complex messages for determining the information contents to apply to mobile services finally.
For the conciseness of description, we will use the notation in Table 1.
4.1. Basic Context Types
The context-aware system would like and is able to know many basic context types via sensors
laid out in mobile communication systems and personal mobile devices. The proposed
mechanism utilizes the values of basic context types as input, which includes background noise,
strength level of positioning signal, user’s moving speed, working circumstance, and usage
frequency of device. A set U is used to indicate these basic context types as follows:
U = {u1, u2, u3, u4, u5} (1)
MS
C
Context
server
BSC
Cel
l
BS
M
S

32
Table 1. Definition of notations
Notation Description
u1 Background noise
u2 Level of positioning signal
u3 User’s moving speed
u4 Working circumstance
u5 Usage frequency of device
v1 Low membership degree
v2 Moderate membership degree
v3 High membership degree
W1 Weight sets for location
W2 Weight sets for movement
W3 Weight sets for user task
W4 Weight sets for user situation
4.2. Membership Functions
In addition to the input values, each basic context type has its own corresponding membership
function. Except the pure number of membership degree, it is important to pay more concern
about the relation between all membership degrees.
Figure 2. Membership function µ1(x) for background noise
Standardization is an important step with membership function, which can avoid the effect of the
extreme value no matter too large or small. For example, the membership function of background
sound µ1(x) is shown in Figure 2. The high, moderate and low membership degrees are defined
as v1, v2 and v3 respectively in light of every membership function. The fuzzy set V formed by
them is shown as follows:
V = {v1, v2, v3} (2)
4.3. High-Level Context Types and Weights
The high-level context types are derived from basic context data, which are the processing result
and outputs from the proposed mechanism. For high-level context type, a user’s situation could
be classified into four categories of characteristics. For instance, the second one is moving speed.

33
The possible states include still, walk, running, bicycle and fast vehicle. Each category of output
has its own weight set. As shown in a weight set Wi, x1 represents the related level indicating the
relationship for u1. The second one is x2 representing the corresponding level for u2 (i.e., the
relationship for u2). The rest can be deduced by analogy, i.e. xk for uk, k=1, 2, 3, 4, 5. In other
words, weights are determined based on the levels of relationship between set U and high-level
context types. Hence, Wi is defined as formula (3).
Wi = {x1, x2, x3, x4, x5} where i = 1, 2, 3, 4 and x = a, b, c, d (3)
Besides, the corresponding formulas for each high-level context type are shown as follows:
W1 = {a1, a2, a3, a4, a5} (3.1)
W2 = {b1, b2, b3, b4, b5} (3.2)
W3 = {c1, c2, c3, c4, c5} (3.3)
W4 = {d1, d2, d3, d4, d5} (3.4)
4.4. Fuzzy Logical Operation
To start with the set U and set V can be now gained from the definitions of formula (1) and (2).
Secondly, the matrix R is listed for the following operation step:
The matrix element rkj represents a value which some basic context element uk (e.g., u1)
corresponds with some moderate membership degree vj (e.g., v2) in the membership function. A
composition operation on fuzzy relations is performed, which aims to associate the related sets. It
operates with a given fact to produce an output which makes the decision in a fuzzy way. The
operative formula is as follows:
4.5. Outputs
The three formulas (3), (4) and (6) can be used to compute and know some characteristic like user
situation whenever we need to. Let the set Wi make a composition with matrix R accordingly,
and then an output set is generated:
Oi = Wi×R

34
= [o11 o12 o13] (6)
Hence, we can easily get the vectors Oi through the computation. As if some mobile nodes have
the similar oij, the users of these devices will be considered engaging in the similar/same activity
or situation. More details will be shown on the strength of those examples in Section 5.
Most important, the classification can be obtained through the mechanism. The rule base shown
in Table 2 is necessary to conclude the classification. Therefore, the classification result about
various characteristics can be generated. The parameter oij will be in connection with the
composition of inputs. Because they conform to certain tendencies, the classification can be
realized. In addition, a prototype of rule base is built up and the adjustment of classification
could be emended via the user feedback when a system is implemented.
The value of parameter oij is connected with the composition of inputs. For instance, suppose W2
= {0.2, 0.5, 0.8, 0, 0} is fitting to a crowd scenario. As the parameters of O2 fits in with the
following relationship: o13>o12 ∧ o12>o11, it indicates that the tendency level for background
noise, user’s moving speed, and usage frequency of device is strong. Thus the user might be
moving at high speed by taking a vehicle. If all weight set Wi and R are known, the full result O
= {O1, O2, O3, O4} can be produced by simple computation.
Table 2. The Rule Base
Types Logical expression Location (O1) Movement (O2) User task (O3) User situation (O4)
I o11≥o12 ∧ o12≥o13 deep indoor still deep rest very fine
II o11≥o13 ∧ o13>o12 indoor walk rest fine
III o12>o11 ∧ o11>o13 half-indoor fast walk available ordinary
IV o12>o13 ∧ o13>o11 half-outdoor run active complex
V o13>o11 ∧ o11>o12 open vehicle busy disorder
VI o13>o12 ∧ o12>o11 very open fast vehicle extremely busy extremely disorder
4.6. Training
By taking advantaging of training data, excellent weights as relying on the weight decision
approach can be generated. In order to estimates the similarity, the similar degree is innovated,
whose definition is shown as formula 7. Both A and B represent fuzzy sets. The notations ● and ⨁
means inner product, and outer product respectively.
Suppose that some output Oi = [o11, o12, o13] and matrix R are known, it is possible to draft n
kinds of probable weight sets (Wi)t, t = 1~n. Each of them does fuzzy composition with R,
respectively. Then, we acquire (Oi)t and compute the approaching degree between (Oi)t and Oi.
The approaching degree 1 means most similar. Oppositely, degree 0 represents most dissimilar.
A weight set with highest approaching degree which is the best weight set candidate should be
selected.
5. NUMERICAL ANALYSIS
The numerical examples to illustrate how the proposed scheme works are shown in the section.
Thirty students are observed for their work schedule and usage frequency of device in the

35
example. The strength level of positioning signal and background noise are referred to the real
values measured in the campus and its neighbourhood. The range of user’s moving speed for
walking and vehicle are measured respectively.
Figure 3: Membership function µ2(x) for level of positioning signal
Figure 4: Membership function µ3(x) for user’s moving speed
Figure 5: Membership function µ4(x) for working circumstance
The membership functions from µ1(x) to µ5(x) are designed accordingly and shown in Figure 2
to Figure 6. It should be noted that the usage frequency of device expresses the minutes in the
latest one hour. Harikesh et al. demonstrated how the membership function is generated from
the realistic numerical data [12].
5.1. Examples for classification
A numerical example is illustrated in this subsection. For the characteristic classification, we set
up weight sets as follows:
 Background weight sets W1 = {0.1, 0.8, 0.1, 0, 0}
 Movement weight sets W2 = {0.2, 0, 0.8, 0, 0.5}
 User task weight sets W3 = {0.2, 0, 0.1, 0.8, 0.3}

36
 User situation weight sets W4 = {0.7, 0, 0.3, 0.1, 0.2}
Suppose that we want to know the background around a user j. To start with, the proposed
scheme computes the composition of W1 and Rj, i.e. O1. The following steps can be deduced by
analogy. As a result, the resulting output is ORj = {O1, O2, O3, O4}. For example, the complete
data of five mobile nodes are given in Table 3.
Table 3: Numerical examples of 5 mobile nodes
u1 (db) u2 (%) u3 (km/hr) u4 (o’clock) u5 (min)
MS1 80 40 16 16 15
MS2 40 40 10 12 10
MS3 80 70 36 20 5
MS4 20 20 0 22 0
MS5 50 60 5 8 0
O1 of the mobile node MS1 is illustrated as an example. The variable u1 is 80 decibel, so
membership degree v1 is 0, v2 is 0, and v3 is 1 as corresponding to the membership function of
background noise. The rest can be gained in the same reason. The matrix R1 can be gained from
those five membership functions and then compute O1 by formula (6). This detail of operation is
shown as follows:
MS1'
s O1 = W1×R1
= [0.42 0.86 0.1] (8)
As above, the outcome shows that o11 is 0.42, o12 is 0.86, and o13 is 0.1. Consequently, the logical
expression of user’s background are o12>o11 ∧ o11>o13. According to the rule base in Table 2, the
estimate result is the “half-indoor”. Accordingly, run, extremely busy, and extremely disorder are
the estimate results of O2, O3 and O4, respectively. To sum up, the overall result O is as follows.
O = {III, IV, VI, VI} = {half-indoor, run, extremely busy, extremely disorder} (9)
The result O vector reveals that the user with MS1 is running at half-indoor environment. In
addition, he/she is estimated to be busy and has disorder state now. For campus study, the student
may be attending P. E. Class.
The example of a single mobile node is illustrated as above. We proceed to discuss the multi-
nodes example. If a mobile user has a disorder tendency, the entertainment message may not be
suitable for him/her at present. The mobile service provider will be able to know such situation
by pay attention to the user situation (O4). Supposed mobile node MS3 is taken for instance.
Through the process of the membership functions and formula (6) discussed above, the results
are generated as following.
MS3'
s O4 = W4×R3

37
= [0.2 0.06 1] (10)
All results about five mobile nodes are as follows:
MS1'
s O4 = [0.2 0.3 0.7], (11.1)
MS2'
s O4 = [0.3 0.5 0.33], (11.2)
MS3'
s O4 = [0.2 0.06 1], (11.3)
MS4'
s O4 = [0.7 0 0], (11.4)
MS5'
s O4 = [0.3 0.7 0.5]. (11.5)
Both of MS1 and MS3 have the lower o11 value and higher o13 value. They are considered to be
in a similar situation. Additionally, the lower o11 value and higher o13 value imply that the users
of MS1 and MS3 are the disorder tendency, so the entertainment advertising messages are not
suitable for them and will be avoided to be sent in the present. It also can be estimated that the
mobile user of MS4 may be at a fine situation without disturbance.
5.2. Illustration for Weight Decision
The weight decision method has been introduced in Section 4.3. If training data are available, we
can use them to get the weight which is more reliable. For instance, the movement (O2) is by
vehicle and O2 is defined as [0.3 0.2 0.8] which conforms to the logical expression o13>o11 ∧
o11>o12. In addition, a matrix R3 for MS3 has been obtained as follows.
Suppose that there are three weight alternatives for W2, which are {0.2, 0, 0.8, 0, 0.5}, {0, 0.1,
0.8, 0.1, 0.1}, and {0.3, 0, 0.4, 0.2, 0.6}. The three vectors are used to do fuzzy composition
with R. Thus, the computing results are [0.5 0.16 0.76], [0.1 0.16 0.73], and [0.6 0.08 0.78]
accordingly. Because [0.1 0.16 0.73] does not conform to the logical expression o13>o11 ∧ o11>o12,
{0, 0.1, 0.8, 0.1, 0.1} should not be considered. The approaching degree between (Oi)t and Oi can
be computed via formula (7). These gained results are 0.85 and 0.97 respectively. The degree
0.97 is greater than 0.85. Hence, the weight set candidate {0.3, 0, 0.4, 0.2, 0.6} should be
selected.
5.3. Applied Scenarios
Our research intends to discover more facts about users’ surroundings and behavior. This
proposed mechanism in the research is supposed to be applied to some coming mobile
applications, such as mobile advertising. The mobile advertising has two major types including
the pull marketing and the push marketing [13]. Both of them can adopt the information which is
extended from our mechanism. Through the information derived from our mechanism, users’
real-time needs can be probably explored and satisfied instantaneously. This research support the
activity recognition required in the mobile context-aware services, which will help mobile

38
service providers to give the proper services smartly. As you can imagine, it is useless to push the
gas station message to a user walking outdoors. On the contrary, it may be suitable to show him
the coupon of the ice cream storefront.
In addition, the context-aware server will be able to collect real-world context data from multiple
users in the scheme. Based on the analysis of these collected data, it becomes possible to develop
a more reliable high-level context recognition mechanism. In other words, the service provider
will be capable to distribute more accurate context-aware information and applications to mobile
users. Moreover, the trustworthy weight sets can be decided by the training data and weight
decision approach during implementation phase. Finally, the fuzzy logic approach owns features
making it a pretty good performance for chaotic decision problems e.g. activity recognition with
no need of precise, noise-free inputs. Hence, the proposed mechanism will make the context-
aware applications more feasible andflexible.
6. CONCLUSIONS
The proposed scheme contributes a feasible way to discover the data of mobile device users and
explore their surroundings. A fuzzy-based model is used to implement such application, because
it can summarize the data from sensors to reason the users’ contexts. The signal processing
techniques and classification mechanism are also investigated. As a result, it will become
possible to provide the real-time context-aware application to users by their situations
accordingly.
One of the important challenges of providing mobile services is the accurate recognition of
human activities. The technology enables service providers to determine a potential customer's
activity and provide the appropriate context-aware services to him. A technical mechanism for
the realization of the context awareness is provided in this paper. Furthermore, it can be utilized
to recognize the basic information for inferring the characteristic identical or close to a user’s
situation. As long as other important mobile technologies progresses maturely, many amazing
mobile services enhancing mobile commerce will become feasible and available. For instance,
mobile users not only receive the instant promotion message as his are near to a department store,
but also the interesting digital information at a tourist attraction, exhibition and etc.
To sum up, a context-aware mechanism serving an important part of providing mobile services is
studied in this research. A reasoning scheme is discussed to illustrate how to estimate the users’
situations analytically. Thus, more and more applications that satisfy users’ immediate need
could be developed in the near future, according to the concept revealed in this paper. Besides,
based on the proposed mechanism, novel mobile services will have more diversity and offer
more chance to enhance mobile commerce with feasibility and flexibility.
REFERENCES
[1] Omer B. Sezer, E. Dogdu, and Ahmet M. Ozbayoglu (2018) Context-Aware Computing, Learning,
and Big Data in Internet of Things: A Survey, IEEE Internet of Things, vol. 5, no. 1, pp. 1–27
[2] Petcovici A. and Stroulia E. (2016) Location-based services on a smart campus: A system and a
study, Proceedings of IEEE the 3rd World Forum on Internet of Things, pp. 94–99
[3] Marek R. Ogiela and Lidia Ogiela (2013) Cognitive Information Systems in Ubiquitous Computing,
Proceedings of the 7th International Conference on Innovative Mobile and Internet Services in
Ubiquitous Computing, vol. 43, no. 2, pp. 246–248
[4] P. Korpipaa, J. Mantyjarvi, J. Kela, H. Keranen, and E.J. Malm (2003) Managing Context
Information in Mobile Devices, IEEE Pervasive Computing , vol. 2, no. 3, pp. 42–51
[5] M. Mühlenbrock, O. Brdiczka, D. Snowdon, and J.-L. Meunier, (2004) Learning to Detect User

39
Activity and Availability from a Variety of Sensor Data, Proceedings of the Second IEEE Annual
Conference on Pervasive Computing and Communications, pp. 13–22
[6] Gorka Azkune and Aitor Almeida (2018) A Scalable Hybrid Activity Recognition Approach for
Intelligent Environments, IEEE Access, vol. 6, pp. 41745–41759
[7] F. Nocera, M. Mongiello, A. Parchitelli, E. Di Sciascio, and L. Patrono (2019) A model for Reflective
Middleware based on fuzzy rule for context-awareness injection in ubiquitous computing
environments, Proceedings of 4th International Conference on Smart and Sustainable Technologies,
pp. 1–7
[8] Liqin Guo, Lixue Wang, Yongyi He, and Sheng Bao (2017) Navigation System of the Stacking
Vehicle Based on Fuzzy Control and Laser Scanner, Proceedings of 10th International Symposium on
Computational Intelligence and Design, vol. 2, pp. 97–102
[9] Sadik K. Gharghan, R. Nordin, Aqeel M. Jawad, Haider M. Jawad, and M. Ismail (2018) Adaptive
Neural Fuzzy Inference System for Accurate Localization of Wireless Sensor Network in Outdoor and
Indoor Cycling Applications, IEEE Access, vol. 6, pp. 38475–38489
[10] Dianshuang Wu, Guangquan Zhang, and Jie Lu, (2015) A Fuzzy Preference Tree-Based
Recommender System for Personalized Business-to-Business E-Services, IEEE Transactions on
Fuzzy Systems, vol. 23, no. 1, pp. 29–43
[11] Al-Qerem Ahmad, (2014) "Performance Evaluation of Transaction Processing in Mobile Data Base
Systems", International Journal of Database Management Systems, vol. 6, no. 2, pp. 19-38.
[12] Harikesh B. Yadav and Dilip K. Yadav, (2015) "A method for generating membership function from
numerical data", Journal of Intelligent & Fuzzy Systems, vol. 29, pp. 2227, 2015.
[13] Ruibing Wang, Qinglong Gou, Tsan-Ming Choi, and Liang Liang (2018) Advertising Strategies for
Mobile Platforms With “Apps”, IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol.
48, no. 3, pp. 757–778
AUTHORS
Kuen-Liang Sue received the M.S. degree in computer science and information
engineering and the Ph.D. degree in information management from National Chiao Tung
University, Hsinchu, Taiwan. He is currently an assistant professor with the Department of
Information Management, National Central University, Zhongli, Taiwan. His research
topics include mobile computing, IoT, and network security, etc.

A Context-aware Strategy for Mobile Services in M-commerce

More Related Content

Similar to A Context-aware Strategy for Mobile Services in M-commerce (20)

More from ijdmsjournal (20)

Recently uploaded (20)

A Context-aware Strategy for Mobile Services in M-commerce