32
ACES Newsletter 2019 - University of Peradeniya
ACES Newsletter 2019 - University of Peradeniya
purchasing habits in stores to determine the
placement of products accordingly. MerchX
is a software platform to give merchandising
suggestions in order to increase the revenue
of shopping malls. It is made as a scalable
tool that can be used as both commercial
software that is helpful for retailers and
also as a data mining tool for future research
purposes.
Detecting Dengue Spreading in Sri Lanka
based on News Articles
Nishara Kavindi, Prabhashi Meddegoda, Peshali
Randika
Improvements for Existing Computer Supervisors: Mr. D.S.Deegalla
Adaptive Systems with Efficient Question The use of indicator-based surveillance
Classification Techniques
systems is the traditional approach of
W.M.K.D. Abeysinghe, K.L.D. Deshapriya, W.A.M.N. monitoring diseases, which use structured
data. The use of event-based surveillance
Weerasooriya
systems is the modern approach, where
Supervisors: Mr. D.S.Deegalla, Prof. Roshan G. Ragel
unstructured data such as information from
Smart learning and adaptive learning are the internet and social media are used. The
replacing the traditional methods of learning. aim of this research was to implement an
This research focused on the improvements automated event based system which query
to existing adaptive testing systems in for newly published online news articles
terms of a Moodle Adaptive Quiz Plugin by and classify them as Dengue-related or not,
integrating question classification with it. extract useful information out of Dengue-
Questions were classified based on subject related articles about Dengue outbreaks in
areas and difficulty levels using different Sri Lanka and store them in a database and
machine learning techniques.
visualize through a web application.
MerchX - A Scalable Software Platform to
Give Merchandising Suggestions for Retail
Industry
J.A. Upeksha Iwanthi, D.A. Dilki Mahindika, E.L.
Nimodi Dilka Madubhashini
Supervisors: Mr.D.S.Deegalla, Dr. Suneth Namal
Karunarathna, Dr. Asitha Bandaranayake
This research was based on analyzing
correlations, new trends, and customers’
an optimization scheme based on advised
reinforcement learning.
Neuro-Fuzzy Dynamic Difficulty Adjustment
for Computer Games
Theekshana Dissanayake, Yasitha Rajapaksha, Heshan
Sandeepa
Supervisors: Prof.Roshan G. Ragel, Dr. Isuru Nawinne
Context-Aware Selective Optimization in
Wireless Mesh Networks with Artificial Dynamic Difficulty Adjustment (DDA) in
Intelligence
computer games is a relatively new research
area which focuses on improving the gaming
Samurdhi Karunarathne, Kalana Suraweera
experience by adjusting the difficulty level of
Supervisors: Dr. Asitha Bandaranayake
the game depending on user performance.
IEEE 802.11 (Wi-Fi) network access has This research focused on the design and
become so ubiquitous in recent years that implementation of a performance-based
one expects such connectivity everywhere, DDA system which employs a combination of
whether at home, workplace, restaurant or a neural network and a fuzzy system. In this
plane. Due to their poor coverage and low system, a multilayer perceptron (MLP) neural
QoS guarantees, single access point (AP) network acts as the difficulty detector and a
networks have failed to meet increasing fuzzy system acts as the difficulty adjuster.
broadband service requirements, resulting Both these components were integrated into
in a demand for multi-AP networks called a first-person shooter game developed from
Wireless Mesh Networks (WMNs). A WMN scratch. The MLP neural network developed
generally consists of mesh gateways (MGs), uses the game parameters to predict the
mesh routers (MRs), mesh clients (MCs) difficulty level of the next section, and then
and a set of wireless links among them. A feed the results into the fuzzy engine. After
network request originating at an MC would that, the fuzzy engine incorporates the game
be transferred through its associated MR onto state parameters and the difficulty value
the wireless backbone, where it takes one or with the knowledge base and computes an
more hops to reach an MG before reaching adequate difficulty adjustment.
the Internet (and vice versa). This research
identified that there is something amiss in
past work that has attempted to increase
throughput or reduce delay for the average
user. It may be sub-optimal to optimize
either throughput or delay for all clients in
a generalized fashion; enhancing throughput
for throughput-demanding clients and
reducing delay for delay-sensitive clients may
lead to much better-perceived performance.
In this work, they have ventured to find such
33