Multitasking in Divided Attention

Bachelor thesis
by
Bhanu Pratap Singh Bankoti
Chandra Shekhar Gupta

Under the supervision of

Dr. Oishila Bandyopadhyay

A thesis submitted to
Indian Institute of Information Technology, Kalyani

for the partial fulfillment of the degree of

Bachelor of Technology
in
Department of Computer Science & Engineering

Spring Semester - 2019

 Declaration

We hereby declare that the work being presented in this thesis entitled,
“Multitasking in Divided Attention”, submitted to Indian Institute of Infor-
mation Technology, Kalyani in partial fulfillment for the award of the degree
of Bachelor of Technology in Computer Science and Engineering during the
period from Jan, 2019 to May, 2019 under the supervision of Dr. Oishila
Bandyopadhyay, Department of Computer Science and Engineering, Indian
Institute of Information Technology, Kalyani, West Bengal 741235, India,
does not contain any classified information.

Bhanu Pratap Singh Bankoti (39/CSE/15014/69)

Chandra Shekhar Gupta (39/CSE/15015/70)
Department: Computer Science & Engineering
Institute: Indian Institute of Information Technology, Kalyani

This is to certify that the above statement made by the candidate is

correct to the best of my knowledge.

Signed:

Date:

i
 Certificate

This is to certify that the thesis entitled “Multitasking in Divided Atten-

tion” being submitted by Bhanu Pratap Singh Bankoti (39/CSE/15014/69)
and Chandra Shekhar Gupta (39/CSE/15015/70), an undergraduate student
in the Department of Computer Science and Engineering, Indian Institute
of Information Technology, Kalyani, West Bengal, India, for the award of
Bachelor of Technology in Computer Science and Engineering is an original
research work carried by them under my supervision and guidance. The the-
sis has fulfilled all the requirements as per the regulations of Indian Institute
of Information Technology, Kalyani and in my opinion, has reached the stan-
dards needed for submission. The work, techniques and the results presented
have not been submitted to any other University or Institute for the award
of any other degree or diploma.

Dr. Oishila Bandyopadhyay

Assistant Professor
Department of Computer Science and Engineering
Indian Institute of Information Technology, Kalyani

Date:
 Acknowledgments

We would like to express our sincere thanks and gratitude to our Super-
visor Dr. Oishila Bandyopadhyay, who gave us the golden opportunity to do
this wonderful project and guided us immensely through the course of the
project. We came to know about so many new things, all thanks to her.

We are also thankful to Prof. Mallika Banerjee, Director of Pradip Centre

for Autism Management, Kolkata, who helped us a lot in doing this project
and we learnt a lot from her.

We also thank our parents for their motivation and support. We must
thank to our classmates for their timely help and support for completion of
this project.

Last but not the least, We would like to thank all those who had helped
directly or indirectly in this project.

Bhanu Pratap Singh Bankoti (39/CSE/15014/69)

Chandra Shekhar Gupta (39/CSE/15015/70)
Indian Institute of Information Technology, Kalyani

Date:

iii
 Abstract

Assessment of cognitive functionality is an important aspect of care for

people. Unfortunately, few tools exist to measure divided attention, the
ability to allocate attention to different aspects of tasks. An accurate deter-
mination of divided attention would allow inference of generalized cognitive
decline, as well as providing a quantifiable indicator of an important com-
ponent of driving skill. Divided attention in vision is fundamentally about
the dependence versus the independence of visual processing across stimuli.
We propose a new method for determining relative divided attention ability
through unobtrusive monitoring of computer use. Specifically, we measure
performance on a multi-task cognitive computer exercise as part of our test
for Divided Attention. This metric indicates whether the user has the ability
to pay attention to all the tasks at once, or is primarily attending to one task
at a time (sacrificing optimal performance). The monitoring of divided at-
tention in a home environment is a key component of both the early detection
of cognitive problems and for assessing the efficacy of learning patterns.

Keywords: Divided Attention, Optimal Performance, unobtrusive mon-

itoring, cognitive computing.

iv
Contents

Acknowledgements iii

1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Types of attention . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Determinants of attention . . . . . . . . . . . . . . . . 2
1.2 Divided Attention . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Attention Assessment 4
2.1 Current methods of assessment . . . . . . . . . . . . . . . . . 4
2.2 Attention Test . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Test for Divided Attention . . . . . . . . . . . . . . . . . . . 5

3 Work Done 6
3.1 Proposed algorithm . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.2.1 Task 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2.2 Task 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.3 Task 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.3 Labelling of Dataset . . . . . . . . . . . . . . . . . . . . . . . 11

4 Classification model and Results 14

4.1 Classification of Divided Attention . . . . . . . . . . . . . . . 14
4.1.1 Naive Bayes Classifier . . . . . . . . . . . . . . . . . . 14
4.1.2 Support Vector Machine (SVM) . . . . . . . . . . . . . 15
4.1.3 Random Forests . . . . . . . . . . . . . . . . . . . . . . 16
4.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

5 Analysis on basis of divided attention 21

5.1 Trends in academic performance . . . . . . . . . . . . . . . . . 21
5.2 Analysis of Linear Perception and Divided Attention . . . . . 22

v
6 Discussion 24

vi
Chapter 1

Introduction

1.1 Background
Attention, in psychology is the concentration of awareness on some phe-
nomenon to the exclusion of other stimuli. Attention is best described as
the sustained focus of cognitive resources on information while filtering or
ignoring extraneous information. It is a very basic function that often is a
precursor to all other neurological/cognitive functions. It is awareness of the
here and now in a focal and perceptive way. Although human experience is
determined by the way people direct their attention, it is evident that they
do not have complete control over such direction. There are, for example,
times when an individual has difficulty concentrating attention on a task, a
conversation, or a set of events. At other times an individual's attention is
captured by an unexpected event rather than voluntarily directed toward it.

1.1.1 Types of attention

Focused attention: The ability to respond discretely to specific visual, au-
ditory or tactile stimuli.
Sustained attention (vigilance and concentration): The ability to main-
tain a consistent behavioral response during continuous and repetitive activ-
ity.
Selective attention: The ability to maintain a behavioral or cognitive set
in the face of distracting or competing stimuli. Therefore, it incorporates the
notion of “freedom from distractibility".
Alternating attention: The ability of mental flexibility that allows in-
dividual's to shift their focus of attention and move between tasks having
different cognitive requirements.
Divided attention: This is the highest level of attention and it refers to the

1
ability to respond simultaneously to multiple tasks or multiple task demands
[12].

1.1.2 Determinants of attention

• Objective features (inherent to external object)

– Motivation
– Intensity
– Size
– Change
– Color
– Clarity
– Repetition

• Subjective features (inherent to individuals):

– Interest
– Motives
– Mental set
– Emotional state
– Habit

1.2 Divided Attention

Divided attention could be defined as our brain’s ability to attend to two
different stimuli at the same time, and respond to the multiple demands of
your surroundings [5]. Divided attention is a type of simultaneous attention
that allows us to process different information sources and successfully carry
out multiple tasks at a time. This cognitive skills is very important, as it
allows us to be more efficient in our day-to-day lives. Our ability to attend
to multiple stimuli and do various tasks at a time does have its limits. When
you divide your attention, the efficiency with which you do these actions is
decreased, and you will almost certainly perform poorly. Interference is the
term used to describe when a person has a hard time attending to two stimuli
at a time. We see interference when the brain is only able to process a cer-
tain amount of information. However, cognitive training can help improve
divided attention, and as a consequence, the ability to do more than one

2
activity at a time [12].

Examples of divided attention

• Divided attention is an important factor in the academic setting. Being

able to understand what the teacher is saying while reading the board
and taking notes are tasks that are essential to successful learning.
This may be one of the reasons why people with attention disorders
(like ADHD) perform poorly in school [8].

• A truck driver is driving on the highway and starts overtaking someone.

As they are passing, they see a sign for their exit. If the driver isn't
able to safely pass the other car and pay attention to traffic signs, they
may lose important information, or even cause danger to themselves
and others. Divided attention is very important to driving safely and
successfully [1] [4].

• When you are eating and talking at the same time, or even when you are
watching TV and talking on the phone, you are using divided attention.

Like most cognitive abilities, divided attention typically increases during

childhood, plateaus in the years following puberty, and then gradually de-
creases with continued aging. The decline in this cognitive skill is one of the
primary facets of decreased driving ability in the elderly [2].

The loss of divided attention skill is also highly associated with other age-
related risks [3], including falls and the onset of Mild Cognitive Impairment
(MCI) [7], often a precursor to Alzheimer's Disease. Screening tests for MCI
are coarse at best; by the time someone does poorly enough to be considered
impaired, his or her family have often been worried for months or years. This
is especially problematic, considering that current pharmacological agents for
Alzheimer's treatment can only slow the symptoms of the disease, not bring
back what has been lost.

3
Chapter 2

Attention Assessment

2.1 Current methods of assessment

Given the importance of divided attention ability, some way to measure it,
with finer resolution than simply being able to declare someone cognitively
impaired or unsafe, would be invaluable for both early detection of problems
and for monitoring progress on cognitive interventions.
Assessing divided attention may be helpful in professional areas, where
divided attention is key (drivers, athletes, etc.). It can also help in academic
fields (if a student needs extra time taking notes or completing certain tasks),
or clinical areas (maybe a patient needs more time to collect the proper
information). In all of these areas, a cognitive assessment [6], may directly
help the user understand their day-to-day lives in a more in-depth context.
Our approach to monitoring cognitive skills like divided attention consists
of embedding algorithms to measure aspects of cognitive performance into
self reinforcing (i.e., enjoyable) adaptive computer games that are played
in the home. This frequent monitoring allows us to trend both absolute
performance and variability in performance with unobtrusive measures. This
technique also lends itself to diagnosis and management of other neurological
difficulties, such as children with autism [10].

2.2 Attention Test

Usually an attention test requires to detect something in a lot of stimuli. For
example, searching 4 dots in lines of groups of 3 or 5 dots, or detecting a 4
between numbers (0-9) sequentially displayed on a computer screen. In such
tests you just have to focus on one task or thing and try not to be distracted
by another (distractor) task. These attention tests are set up to investi-

4
gate concentration (see my page on Attention problems for an explanation
of different types). Unfortunately, attention tests rarely resemble daily life
activities and it is therefore questionable how accurately they really measure
your concentration skills in daily life. One thing is certain though, most
such tests require quite a lot of focus. Probably much more than daily life
activities.

2.3 Test for Divided Attention

It is the attention tests developed to measure divided attention by doing more
than two things at once. There are not many tests around that assess this
aspect of attention. Here, we attempt to formalize a method for analyzing
one game which is designed to challenge player's ability to attend to multiple
dimensions of stimuli simultaneously.
We have created a game consisting of 3 tasks in which a user has to
simultaneously response to all 3 tasks in a sequential order. This game runs
for around 12 minutes.

5
Chapter 3

Work Done

We have introduced visual search and multitask paradigm to study the di-
vided attention. Initial consideration of divided attention is limited to a
consideration of simple stimuli and tasks and focuses on divided attention
across space.

3.1 Proposed algorithm

We will show divided attention effects, performance will depend on the num-
ber of task-relevant stimuli. A contrasting extreme to serial processes would
be those that proceed simultaneously (i.e., in parallel) for multiple stimuli
and are unaffected by the number of to-be-processed stimuli. For some situ-
ations, a set of completely independent parallel process is expected to suffer
no divided attention effects. Besides serial versus parallel processing, there
are other types of dependencies that can lead to divided attention effects [9].
Effects of divided attention are behavioral consequences (such as impaired
performance) of manipulating the number of relevant stimuli. We offer an
analysis, asking what aspect of processing is the source of the divided atten-
tion effects. In this test we are given a randomly generated stimuli which
keeps on changing and we have to respond for the target stimuli. The purpose
is to introduce the paradigms, recognize some of their strengths and weak-
nesses, and understand the attention level that can and cannot be addressed.

To classify the divided attention in 3 category namely Low, Average and

High divided attention, we have applied the following steps after collecting
the data:

• Remove the inconsistent data from dataset.

6
 • Label the data based on the mean and standard deviation of the at-
tention score.

• Apply machine learning model to classify the attention type (Low, Av-
erage, High).

• Analyze the output and trends in academic performance and divided

attention.

3.2 Function
In this work we have developed three distinct task which will have spatial
identification test, color and number test (Figure 3.1), which will monitor
the attention level by using visual effect and multi-tasking. A user has to
press the specified key at instant, the target stimuli occur [11].
In this test, there is predefined input which will keeps on changing randomly
at the time interval of 750 milliseconds. This whole process of test occur
in order i.e., firstly angle determination test is done (task 1), then color se-
lection test (task 2) and then number selection test (task 3) (Figure 3.1).
Each test is of 3 seconds and this is repeated 20 times. If a user is unable
to respond within that 6 seconds, either he missed it or unable to act, it will
be considered as wrong outcome. If user is fast and smart enough to act
decisively and press right key for that target stimuli, answer is recorded as
true, otherwise false value is stored in database.

Figure 3.1: Window containing all 3 task.

7
3.2.1 Task 1
In first task, we try to measure the attention level of a user related to spatial
movement of stimuli. In this test, an angle is generated randomly and it
keeps on changing at each 750 milliseconds. A user is given input as particu-
lar angle (in degree) and he have to click the “left arrow" at the instant when
that particular angle (target stimuli) is shown in the clock (Figure 3.2). This
task run for 6 seconds only and user has to respond correctly within this 6
seconds. At the same instant, evaluation is done and the output is stored in
database. And this task again run after task 2 and task 3 are performed. 20
records are created in this process.

Figure 3.2: Preview of Task 1.

8
3.2.2 Task 2
In this task, we are concerned with chromatic stimulus behavioural process.
We have to select a given color out of the rest of the colors (8 colors) which
are generated randomly at regular interval of time and changes periodically
at time interval of 750 milliseconds (Figure 3.3). We will use “up-arrow"
for responding for the given color (target stimuli). If the time limit (i.e., 6
seconds for each question) is elapsed and the color is not selected then the
question will be marked missed. For example, in Figure 3.3 target stimuli is
gray and instant stimuli is blue.

Figure 3.3: Preview of Task 2.

9
3.2.3 Task 3
In this game we are concerned with stimulus and response related to number
game. We have to correctly select a number which is generated randomly
using a pseudo number generator that keeps on changing every 750 millisec-
onds (ranging from 1-8 inclusive). We will use “right-arrow" for selecting the
given target color. If the time limit (i.e., 6 seconds for each question) is ex-
ceeded and the number is not yet selected then the question will be marked
missed. This task is also performed 20 times. For example, in given Figure
3.4, a user is asked to press right key when orange square box shows 5 as
target stimuli is 5.

Figure 3.4: Preview of Task 3.

10
3.3 Labelling of Dataset
The dataset that is collected is not labeled so we have done statistical analysis
on the data in order to label it as High, Average or Low divided attention.
The steps for labeling a dataset is as follows :

1. Firstly, we have done the statistical analysis individually for each of the
three cognitive tasks. We are using mean(µ) and standard deviation(σ)
in order to differentiate between the three classes. Let the 3 cognitive
tasks be task-a, task-b, and task-c then the corresponding statistical
parameters for these are (µa , σ a ), (µb , σ b ), (µc , σ c ). We have also clas-
sified the performance in these individual tasks into 3 classes(High(H),
Average(A), and Low(L)) as follows :
Ha = above µa + σ a Aa = between µa - σ a and µa + σ a
La = below µa - σ a
Hb = above µb + σ b Ab = between µb - σ b and µb + σ b
Lb = below µb - σ b
Hc = above µc + σ c Ac = between µc - σ c and µc + σ c Lc
= below µc - σ c

2. Again we will do the statistical analysis but now for the combined
cognitive task comprising of all the three sub-tasks a, b and c. The
statistical parameters for this combined task are (µ, σ).

3. Then we will check for the consistency of the subject's performance in

the combined cognitive task with the help of the performance of his
individual tasks. By doing so all the inconsistent or irrelevant records
(due to fatigue, boredom or habituation) will be filtered out.

4. We will categorize each record of the dataset in three classes of Divided

Attention as follows :

• High Divided Attention (results are above µ + σ)

• Average Divided Attention (results are between µ - σ and µ + σ)
• Low Divided Attention (result are below µ - σ)

11
 Scatter plot of Task-1, Task-2 and task-3 are shown in Figure 3.5, in
which green bubble represent High attention score, red dot represent Average
attention score, whereas blue bubble represent Low attention score.

Figure 3.5: Scatter plot for all 3 tasks

12
 Index Task-1 Task-2 Task-3 Z

0 8 13 13 0

1 13 11 12 0

2 7 10 18 -1

3 8 14 11 0

4 11 16 17 1

5 14 15 12 1

6 9 10 16 0

7 13 12 11 0

8 9 9 15 0

9 8 15 17 1

10 12 13 11 0

11 15 16 15 1

12 15 14 15 1

Table 3.1 Labelled Dataset

Target Value Z (in table 3.1) is divided into 3 class namely -1, 0, 1 where,

• -1 corresponds to Low Divided Attention.

• 0 corresponds to Average Divided Attention.

• 1 corresponds to High Divided Attention.

13
Chapter 4

Classification model and Results

4.1 Classification of Divided Attention

In order to classify the Divided Attention of a subject, we have used some
of the machine learning algorithms. We are using classification which is a
technique for determining class the dependent belongs to based on the one
or more independent variables. Here the classes are High, Average and Low
and the features are the correct responses of the subject on the basis of 3
cognitive tasks. Before applying any classification algorithm we have split
the dataset into two sets, one for training dataset(60%) and the other for
testing data-set(40%) before training our model. Then we are employing
three classification techniques:

4.1.1 Naive Bayes Classifier

A Naive Bayes classifier is a probabilistic machine learning model thatâĂŹs
used for the classification task. The crux of the classifier is based on the
Bayes theorem with the independence assumptions between predictors i.e.,
it assumes the presence of a feature in a class is unrelated to any other fea-
ture. Even if these features depend on each other or upon the existence of
the other features, all of these properties independently. Thus, the name
Naive Bayes [13].

Naive Bayes classifier is the fast, accurate and reliable algorithm. Naive
Bayes classifiers have high accuracy and speed on large datasets. Naive
Bayes classifier assumes that the effect of a particular feature in a class is
independent of other features. Even if these features are interdependent,
these features are still considered independently. This assumption simplifies

14
computation, and that’s why it is considered as naive. This assumption is
called class conditional independence.

P (D|h)P (h)
P (h|D) =
P (D)
• P(h): the probability of hypothesis h being true (regardless of the data).
This is known as the prior probability of h.

• P(D): the probability of the data (regardless of the hypothesis). This

is known as the prior probability.

• P(h|D): the probability of hypothesis h given the data D. This is known

as posterior probability.

• P(D|h): the probability of data d given that the hypothesis h was true.
This is known as the posterior probability.

Naive Bayes classifier calculates the probability of an event in the following

steps:

1. Calculate the prior probability for given class labels.

2. Find Likelihood probability with each attribute for each class.

3. Put these value in Bayes Formula and calculate posterior probability.

4. See which class has a higher probability, given the input belongs to the
higher probability class.

4.1.2 Support Vector Machine (SVM)

A Support Vector Machine (SVM) is a discriminative classifier formally de-
fined by a separating hyperplane. In other words, given labeled training data
(supervised learning), the algorithm outputs an optimal hyperplane which
categorizes new examples. It is based on the concept of decision planes that
define decision boundaries. A decision plane is one that separates between
a set of objects having different class memberships [14]. SVM offers very
high accuracy compared to other classifiers such as logistic regression, and
decision trees.

The main objective is to segregate the given dataset in the best possible
way. The distance between either nearest points is known as the margin. The

15
objective is to select a hyperplane with the maximum possible margin be-
tween support vectors in the given dataset. SVM searches for the maximum
marginal hyperplane in the following steps:

1. Generate hyperplanes which segregates the classes in the best way.

2. Select the right hyperplane with the maximum segregation from either
nearest data points.

4.1.3 Random Forests

Random Forest Classifier is an ensemble algorithm based on bagging i.e.,
bootstrap aggregation. Ensemble methods combine more than one algo-
rithms of the same or different kind for classifying objects i.e., an ensemble
of SVM, Naive Bayes or Decision Trees. Random forests create decision trees
on randomly selected data samples, gets a prediction from each tree and se-
lects the best solution by means of voting [15]. It also provides a pretty good
indicator of the feature importance.

Deep decision trees may suffer from overfitting, but random forests pre-
vent from overfitting by creating trees on random subsets. The main reason
is that it takes the average of all the predictions, which cancels out the biases.
Random Forest adds additional randomness to the model while growing the
trees. Instead of searching for the most important feature while splitting a
node, it searches for the best feature among a random subset of features.
This results in a wide diversity that generally results in a better model.

It technically is an ensemble method (based on the divide-and-conquer

approach) of decision trees generated on a randomly split dataset. This col-
lection of decision tree classifiers is also known as the forest. The individual
decision trees are generated using an attribute selection indicator such as
information gain, gain ratio, and Gini index for each attribute. Each tree
depends on an independent random sample. In a classification problem, each
tree votes and the most popular class is chosen as the final result. In the
case of regression, the average of all the tree outputs is considered as the fi-
nal result. It is simpler and more powerful compared to the other non-linear
classification algorithms.

16
It works in four steps:

1. Select random samples from a given dataset.

2. Construct a decision tree for each sample and get a prediction.

3. Perform a vote for each predicted result.

4. Select the prediction result with the most votes as the final prediction.

17
4.2 Results
In order to classify the Divided Attention of a subject, we have employed
three machine learning classification algorithms (SVM, Naive Bayes classi-
fier and Random Forest). We have done a comparative analysis among them
on the basis of their efficiency. From the analysis we can conclude that the
efficiency of Random forest classifier is better as compared to SVM and Naive
Bayes classifier. Apart from this, we have also plotted the confusion matrix
of all the three classifiers that can be used to describe the performance of
our classification model(or classifier) on a set of test data whose true values
are known.

predicted value

-1 0 1

-1 2 6 0
True value

0 0 49 1

1 0 7 7

Confusion Matrix for Naive Bayes Classifier.

From the above confusion matrix of Naive Bayes Classifier, we conclude

that out of 8 instances of Low attention class, 2 of them are correctly pre-
dicted but 6 of them are incorrectly classified into Average attention class.
For 50 instances of Average class, 49 are correctly classified but 1 instance
is misclassified as High attention class. For 14 instances of High class at-
tention, 7 of them are classified correctly but 7 is incorrectly classified as
Average attention class. So, accuracy of Naive Bayes model is 80.57%.

18
 predicted value

-1 0 1

-1 3 5 0
True value

0 2 48 0

1 0 1 13

Confusion Matrix for SVM Classifier.

From the above confusion matrix of SVM Classifier, out of 8 instances of

Low attention class, 3 instance are predicted correctly but 5 of them are in-
correctly classified into Average attention class. For 50 instances of Average
class, 48 are correctly classified but 2 instances is misclassified as Low atten-
tion class. For 14 instances of High class attention, 13 of them are classified
correctly but 1 instance is incorrectly classified as Average attention class.
So, performance of SVM model is 88.88%.

predicted value

-1 0 1

-1 4 4 0
True value

0 1 49 0

1 0 2 12

Confusion Matrix for Random Forest Classifier.

19
 Above confusion matrix of Random Forest Classifier indicate that for 8
instances of Low attention class, 4 instances are correctly predicted but 4
are incorrectly predicted into Average attention class. For 50 instances of
Average class, 49 are predicted correctly but 1 instance is misclassified as
Low attention class. For 14 instances of High class attention, 12 of them
are predicted correctly but 2 instances is incorrectly predicted as Average
attention class. So, performance of this model for our dataset is 90.27%.
We have found the accuracy of all three classifiers using the given dataset.

The accuracy of SVM classifier is 88.88%, for Naive Bayes classifier it is

80.55% and for Random Forest classifier the accuracy is 90.27% (Table 4.1).
From these statistics we can conclude that Random Forest classifier is more
accurate and hence better classifier for Divided Attention than SVM and
Naive Bayes classification model.

Classifier Accuracy

Naive Bayes 80.55

SVM 88.88

Random Forest 90.27

Table 4.1: Efficiency of implemented classifiers.

20
Chapter 5

Analysis on basis of divided

attention

5.1 Trends in academic performance

This analysis examines the association between student's divided attention
performance and their academic functioning. We will see whether a subject
having high, low or average divided attention performs well or poorly in his
academics. This study helps in examining whether there is a negative asso-
ciation between attention difficulties and their academic performance.

We have collected the data regarding their academic performance from

all the students who performed the test for divided attention. For the anal-
ysis, we have plotted a histogram and scatter plot graph (Figure 5.1). From
these graphs we can analyze that most of the students have average divided
attention in contrast to low and high divided attention. Majority of the
students with high divided attention have good academic performance while
some have average and few of them have poor academic performance. The
students with poor academic performance and high divided attention must
have strong logical and analytical reasoning and high mental agility.

Most of the students with average divided attention have either good or
average academic performance and the students with low divided attention
have poor academic performance.
From these results (Figure 5.1), we can conclude that divided attention has
a strong association with academic functioning and students with attention
problems have a subsequent decline in their academics.

21
 Figure 5.1: Plot of Attention Score and Academic Performance.

5.2 Analysis of Linear Perception and Divided

Attention
All the ways in which we experience the world around us for example, we
recognize our favorite food by its aroma and the way it looks, we recognize
an orange by its round shape, citrus flavor, and its color, etc. It is through
these sensory experiences that we interact with and interpret things in our
world. Recognizing and interpreting sensory information, such as sound and
smells, are all a part of perception.

Perception refers to the way sensory information is organized, interpreted,

and consciously experienced. It also includes how we respond to the informa-
tion. We can think of perception as a process where we take in sensory infor-
mation from our environment and use that information in order to interact
with our environment. Perception allows us to take the sensory information
in and make it into something meaningful.

Here we have designed a linear perception device in which given a line of

some fixed length (L) the subject has to slide a line accordingly to draw the

22
length (L’ ) with minimum error (Error = L’ - L). We are taking the subject's
performance error as a parameter to analyze the relationship between Linear
Perception and their Divided Attention.
For the analysis, we have plotted a graph (Figure 5.2) between Divided
Attention score and Linear Perception response (error).

Figure 5.2: Plot of Attention Score and error in Linear Perception.

From Figure 5.2 we cannot observe any strong association between Lin-
ear Perception and Divided Attention. The possible causes may be due to
insufficient dataset or the discrepancy in the subject's Divided Attention and
most of the data we have collected corresponds to the subject with no at-
tention difficulties. So this analysis can give a better result if we have more
data and more feature parameters (especially data of the subject's facing
with attention problems).

23
Chapter 6

Discussion

Divided attention is our brain's ability to attend 2 or more stimuli simulta-

neously. It allows us to process different information sources and successfully
carry out multiple tasks at a time. Our ability to attend to multiple stim-
uli and do various tasks at a time does have its limits. We have developed
a computer-based, user-friendly attention test for Divided Attention which
consists of three games. Each game is independent and has its own cognitive
requirements. These tests run in sequential order, periodically for about 9
minutes. After the test is completed we use the results of our test as dataset
for our machine learning model. The dataset thus obtained is unlabelled so
we label these dataset into 3 classes of Divided Attention as High, Low or
Average by doing statistical analysis. By doing so all the inconsistent records
are also filtered out. In order to classify the Divided Attention of a subject,
we have employed three machine learning classification algorithms (SVM,
Naive Bayes classifier and Random Forest) and done comparative analysis
among them. We have also examined the association between student's di-
vided attention performance and their academic functioning and established
that students with attention problems have a subsequent decline in their
academics. Apart from this, we have also studied the relationship between
Linear Perception and their Divided Attention by using a linear perception
device. It can be used as a cognitive assessment to measure attention deficits
in patients and normal (healthy) persons.

24
References

[1] Brouwer WH, Waterink W, Van Wolffelaar PC, Rothengatter T. Di-

vided Attention in Experienced Young and Older Drivers: Lane Tracking
and Visual Analysis in a Dynamic Driving Simulator. J Human Factors and
Ergonomics Soc. 1991 Oct;33(5):573âĂŞ582. [PubMed]
[2] Callahan C, Hendrie H, Tierney W. Documentation and evaluation of
cognitive impairment in elderly primary care patients. American College of
Physicians. 1995;122:422âĂŞ429. [PubMed]
[3] Verghese J, Buschke H, Viola L, Katz M, Hall C, Kulansky G, et
al. Validity of divided attention tasks in predicting falls in older indi-
viduals: a preliminary study. Journal of the American Geriatrics Society.
2002;50:1572âĂŞ1576. [PubMed]
[4] Parasuraman R, Nestor PG. Attention and Driving Skills in Aging
and AlzheimerâĂŹs Disease. J Human Factors and Ergonomics Soc. 1991
Oct;33(5):539âĂŞ557. [PubMed]
[5] Scalf PE, Colcombe SJ, McCarley JS, et al. The neural correlates of an
expanded functional field of view. The Journals of Gerontology, Series B,
Psychological Sciences and Social Sciences. 2007;62(1):32âĂŞ44. [PubMed]
[6] Chen P, Ratcliffe G, Phil D, Belle S, Cauley J, et al. Cognitive tests
that best discriminate between presymptomatic AD and those who remain
nondemented. Neurology. 2000;55:1847âĂŞ1853. [PubMed]
[7] Petersen RC, Stevens JC, Ganguli M, Tangalos EG, Cummings JL,
DeKosky ST. Practice parameter: Early detection of dementia: Mild cog-
nitive impairment (an evidence-based review): Report of the Quality Stan-
dards Subcommittee of the American Academy of Neurology. Neurology. May
8;56(9):1133âĂŞ42.[PubMed]
[8] Innes CR, Jones RD, Anderson TJ, Hollobon SG, Dalrymple-Alford JC.
Performance in normal subjects on a novel battery of driving-related sensory-
motor and cognitive tests. Behav Res Methods. 2009;41(2):284âĂŞ94.
[PubMed]

25
[9] Edwards JD, Ross LA, Wadley VG, Clay OJ, Crowe M, et al. The useful
field of view test: normative data for older adults. Arch Clin Neuropsychol.
2006 May;21(4):275âĂŞ86. [PubMed]
[10] Jimison HB, Pavel M. Integrating computerbased health coaching into
elder home care. In: Mihailidis A, Boger J, Kautz H, Normie L, editors.
Technology and Aging. IOS Press; Amsterdam, The Netherlands: 2008.
[11] Jimison HB, Pavel M, Bissell P, McKanna J. A framework for cognitive
monitoring using computer game interactions. Studies in Health Tech and
Informatics. 2007;129(2):1073âĂŞ1077. [PubMed]
[12] http://thepeakperformancecenter.com/educational-
learning/learning/process/obtaining/types-of-attention/
[13] https://www.cc.gatech.edu/ isbell/reading/papers/Rish.pdf
[14] www.jmlr.org/papers/volume2/tong01a/tong01a.pdf
[15] http://www.jmlr.org/papers/volume13/biau12a/biau12a.pdf

26