Barnan Das is a software engineering intern and PhD candidate researching machine learning approaches for elderly care. His research interests include machine learning, smart environments, pervasive computing, and mobile health. He has developed several machine learning techniques to address issues like imbalanced class distributions and overlapping classes in activity recognition data. These techniques have improved automated prompting systems to help elderly individuals with activities of daily living. His work also involves harnessing the sensors and computation of mobile devices to recognize both simple and complex daily activities.

1. Barnan Das
Software Engineering Intern
PC Client Group, Intel
Manager: Narayan Biswal
PhD Candidate
Washington State University
Advisor: Dr. Diane J. Cook

2. Machine
Learning
Smart
Environments
Data
Mining
Research
Interests
Pervasive
Computing
Mobile
Health
2

3. Machine Learning Driven
Caregiving for the Elderly

4. 36
million
Worldwide Dementia
population
13.2m
Actual and expected
number of Americans >=65
year with Alzheimer’s
7.7m
5.1m
2010
2030
2050
$200
Payment for care in 2012
billion
15
Unpaid caregivers
million
4
Source: World Health Organization and Alzheimer’s Association.

5. 5

6. Automated
Prompting
Help with Activities of Daily Living (ADLs)
6

7. Architectural Overview
7
Published at ICOST 2011 and Journal of Personal and Ubiquitous Computing 2012.

8. Experimental Setup
Raw Data
8
daily
activities
150
Sweeping
Cooking
Medication
Watering Plants
Etc.
elderly
participants
Prompts issued
when errors were
committed
Clean Data
1
activity
step
17
1
data
point
engineered features
0/1
Binary class
{prompt, no-prompt}
Length of activity step
Location in apartment
# sensors involves
# distribution of sensor events
Etc.
8

9. Class Distribution
149
Total number
of data points
3980
3831
9

10. Machine Learning
Contribution
Automated
Prompting
Imbalanced
Class
Distribution
Overlapping
Classes
10

11. Machine Learning
Contribution
Automated
Prompting
Imbalanced
Class
Distribution
Overlapping
Classes
11

12. Imbalanced Class
Distribution
12

13. Proposed Approach
 Preprocessing technique to oversample minority class
Approximate discrete
probability distribution using
Generate new minority
class data points using
Chow-Liu’s algorithm
Gibbs sampling
13
Published at International Conference on Data Mining 2013 and IEEE Transaction on Knowledge & Data Engineering 2014

14. (wrapper-based)RApidly COnverging
Gibbs sampler: RACOG & wRACOG
 Differ in generated sample selection
RACOG
wRACOG
Runs for predefined number of
iterations
Stops when there is no further
improvement of the learning model
Effectiveness of new samples is not
judged
Judges effectiveness of new samples
using a Boosting-like method
Total number of new samples
generated is more
Total number of new samples
generated is far less
14

15. Experimental Setup
Datasets
•
•
•
•
•
•
Classifiers
prompting
abalone
car
nursery
letter
connect-4
• C4.5 decision
tree
• SVM
• k-Nearest
Neighbor
• Logistic
Regression
Other
Methods
• SMOTE
• SMOTEBoost
• RUSBoost
Implemented Gibbs sampling, SMOTEBoost, RUSBoost in MATLAB
15

16. Results (RACOG & wRACOG)
Geometric Mean
(TP Rate, TN Rate)
TP Rate
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
16

17. Results (RACOG and wRACOG)
ROC Curve
17

18. Machine Learning
Contribution
Automated
Prompting
Imbalanced
Class
Distribution
Overlapping
Classes
18

19. Overlapping
Classes
19

20. Overlapping Classes in Prompting Data
3D PCA Plot of prompting data
20

21. Tomek Links
21

22. Cluster-Based Under-Sampling(ClusBUS)
Form clusters
Under-sampling
clusters
22
Published in IOS Press Book on Agent-Based Approaches to Ambient Intelligence, 2012 and ICDM Workshop 2013.

23. Experimental Setup
Dataset
prompting
Clustering Algorithm
DBSCAN
Minority class dominance
Empirically determined
threshold
Classifiers
C4.5 Decision Tree
Naïve Bayes
k-Nearest Neighbor
SVM
23

24. Results (ClusBus)
SMOTE
ClusBUS
Original
G-mean
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
C4.5
Naïve
Bayes
IBk
SMO
SMOTE
ClusBUS
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
C4.5
Naïve Bayes
Original
AUC
TP Rate
Original
SMOTE
IBk
SMO
ClusBUS
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
C4.5
Naïve Bayes
IBk
SMO
24

25. Personal and
Pervasive
Sensor
Suite
Computation
Power
25

26. Harnessing Pervasiveness of
Mobile Devices
Locomotive
Activity Recognition
Complex Daily
Activity Recognition
26

27. Harnessing Pervasiveness of
Mobile Devices
Locomotive
Activity Recognition
Complex Daily
Activity Recognition
27

28. Locomotive Activity Recognition
28

29. •
•
•
•
•
•
•
•
Sitting
Standing
Walking
Running
Climbing stairs
Lying
Biking
Driving
Complex
Simple
Activities
•
•
•
•
•
•
Cleaning
Cooking
Medication
Sweeping
Hand washing
Watering plants
29

30. Feature Generation
Sensors
Sampling Rate
Participants
Accelerometer, Rotation Vector Sensor
30 Hz
10
Feature
Acceleration
Rotation Vector
Mean
X, Y, Z
x*sin(/2), y*sin(/2), z*sin(/2)
Min
X, Y, Z
x*sin(/2), y*sin(/2), z*sin(/2)
Max
X, Y, Z
x*sin(/2), y*sin(/2), z*sin(/2)
Standard Deviation
X, Y, Z
x*sin(/2), y*sin(/2), z*sin(/2)
Zero-Crossing Rate
X, Y, Z
Pair-wise Correlation
X/Y, X/Z, Y/Z
30

31. Results: Accuracy
Performance of Different Classifiers
31
Published at International Conference on Intelligent Environments, 2012. [Most Commended Paper Award]

32. Harnessing Pervasiveness of
Mobile Devices
Locomotive
Activity Recognition
Complex Daily
Activity Recognition
32

33. Complex Daily Activity Recognition
Time of
Day
?
Location
Magnetic field-based indoor
location estimation
Simple
Activities
Daily
Activities
Cooking
Eating
Sleeping
Toileting
Brushing
Teeth
Work at Home
Watching TV
Exercising
33

34. Indoor Location Estimation
 Magnetic field along X, Y, Z (T)
 Sampling rate: 30Hz
 50% overlap on sliding window
Bedroom
Bathroom
Kitchen
Dining table
Living room
Living room couch
Home office
Supervised Machine
Learning Model
Location
Prediction
>95%
accuracy
C4.5 Decision Tree
10-fold cross validation
34

35. Performance on Complex Daily Activities
3 weeks
participants
2 apartments
daily
9 activities
Time of day
Accelerometer
Rotation Vector Sensor
Magnetometer  Location
Machine
Learning
Model
Daily Activity
Recognition
>90%
accuracy
C4.5 Decision Tree and kNN
10-fold cross validation
35

36. Conclusion
Algorithms
Applications
Imbalanced
Class
Distribution
Automated
Prompting
Overlapping
Classes
Smart PhoneBased Activity
Recognition
36

37. Publications
Book
Chapters
•
•
•
Journal
Articles
•
•
•
•
Conferences
•
•
•
•
•
•
Workshops
•
•
•
•
B. Das, N.C. Krishnan, D.J. Cook, “Handling Imbalanced and Overlapping Classes in Smart Environments Prompting Dataset”,
Springer Book on Data Mining for Services, 2012
B. Das, N.C. Krishnan, D.J. Cook, “Automated Activity Interventions to Assist with Activities of Daily Living”, IOS Press Book on AgentBased Approaches to Ambient Intelligence, 2012
B. Das, N. C. Krishnan, D. J. Cook, “RACOG and wRACOG: Two Gibbs Sampling-Based Oversampling Techniques”, Transaction of
Knowledge and Data Engineering (TKDE), 2014 (Accepted)
B. Das, D.J. Cook, M. Schmitter-Edgecombe, A.M. Seelye, “PUCK: An Automated Prompting System for Smart Environments”, Journal
of Personal and Ubiquitous Computing, 2012
A.M. Seelye, M. Schmitter-Edgecombe, B. Das, D.J. Cook, “Application of Cognitive Rehabilitation Theory to the Development of
Smart Prompting Technologies”, IEEE Reviews on Biomedical Engineering, 2012
B. Das, N. C. Krishnan, D. J. Cook, “wRACOG: A Gibbs Sampling-Based Oversampling Technique”, International Conference on Data
Mining (ICDM), 2013
S. Dernbach, B. Das, N.C. Krishnan, B.L. Thomas, D.J. Cook, “Simple and Complex Acitivity Recognition Through Smart Phones”,
International Conference on Intelligent Environments (IE), 2012
B. Das, C. Chen, A.M. Seelye, D.J. Cook, “An Automated Prompting System for Smart Environments”, International Conference on
Smart Homes and Health Telematics (ICOST), 2011
E. Nazerfard, B. Das, D.J. Cook, L.B. Holder, “Conditional Random Fields for Activity Recognition in Smart Environments”,
International Symposium on Human Informatics (SIGHIT), 2010
C. Chen, B. Das, D.J. Cook, “A Data Mining Framework for Activity Recognition in Smart Environments”, International Conference on
Intelligent Environments (IE), 2010
B. Das, N. C. Krishnan, D. J. Cook, “Handling Imbalanced and Overlapping Classes in Smart Environments Prompting Dataset”, ICDM
Workshop on Data Mining in Bioinformatics and Healthcare, 2013
B. Das, B.L. Thomas, A.M. Seelye, D.J. Cook, L.B. Holder, M. Schmitter-Edgecombe, “Context-Aware Prompting From Your Smart
Phone”, Consumer Communication and Networking Conference Demonstration (CCNC), 2012
B. Das, A.M. Seelye, B.L. Thomas, D.J. Cook, L.B. Holder, M. Schmitter-Edgecombe, “Using Smart Phones for Context-Aware
Prompting in Smart Environments”, CCNC Workshop on Consumer eHealth Platforms, Services and Applications (CeHPSA), 2012
B. Das, D.J. Cook, “Data Mining Challenges in Automated Prompting Systems”, IUI Workshop on Interaction with Smart Objects
Workshop (InterSO), 2011
B. Das, C. Chen, N. Dasgupta, D.J. Cook, “Automated Prompting in a Smart Home Environment”, ICDM Workshop on Data Mining for
Service, 2010
C. Chen, B. Das, D.J. Cook, “Energy Prediction Using Resident’s Activity”, KDD Workshop on Knowledge Discovery from Sensor Data
(SensorKDD), 2010
C. Chen, B. Das, D.J. Cook, “Energy Prediction in Smart Environments”, IE Workshop on Artificial Intelligence Techniques for Ambient
Intelligence (AITAmI), 2010.
37

38. Barnan Das
 (208) 596-1169
 barnandas@gmail.com
 www.barnandas.com
38