Data Management and Databases Presentation

Data Management & Databases presentation

1. Smart Home Technologies
Data Management and Databases

2. Databases for Smart Homes
Requirements
Database Types
Database Technologies
Smart Home Databases
Data Mining

3. Data Storage Requirements
Sensor data
Temperature (15 @ 8 Kbps)
Humidity (15 @ 8 Kbps)
Gas (15 @ 8 Kbps)
Light (15 @ 8 Kbps)
Motion (15 @ 8 Kbps)
Pressure (100 @ 8 Kbps)
Microphone (15 @ 500 Kbps)
Camera (15 @ 10 Mbps)

4. Data Storage Requirements
User data
Multimedia
Phone messages/conversations (500 Kbps –10 Mbps)
Music (500 Kbps)
TV/Radio broadcasts (500 Kbps –10 Mbps)
Home movies (10 Mbps)
Images
Computer
Programs
Data files
Operating systems

5. Data Storage Issues
Issues
Query frequency and type
Sampling/recording rates
205 sensors (158,900 Kbps)
Multimedia recordings
Simultaneous playback
Analysis, prediction, decision-making queries
Transaction granularity
Historical data, decay
Security and privacy
Centralized vs. distributed

6. What Data to Store
Type of Data
Raw data
Pre-processed
Compressed
Frequency of Data Storage for
Sensor Data
Tradeoff between precision and
quantity

7. Sensor Data Example
9/8/2002 2:0:1 AM~A5 (Coffee Maker) ON
9/8/2002 1:6:59 AM~A9 (A/C) ON
9/8/2002 3:58:52 AM~A0 (Stereo) ON
9/8/2002 5:57:0 AM~A2 (Kitchen Light) ON
9/8/2002 3:1:42 AM~A5 (Coffee Maker) OFF
9/8/2002 7:8:3 AM~A3 (Stove) ON
9/8/2002 12:54:52 PM~A10 (Bathroom Light) ON
9/8/2002 4:58:5 AM~A0 (Stereo) OFF
9/8/2002 8:1:20 AM~A3 (Stove) OFF
9/8/2002 9:6:10 AM~A8 (Computer) ON
9/8/2002 10:8:19 AM~A4 (Bathtub Heater) ON
9/8/2002 11:9:4 AM~A0 (Stereo) ON
9/8/2002 9:4:5 AM~A8 (Computer) OFF
9/8/2002 10:9:4 AM~A4 (Bathtub Heater) OFF
9/8/2002 2:2:5 PM~A10 (Bathroom Light) OFF
9/8/2002 2:52:37 PM~A0 (Stereo) OFF
9/8/2002 4:2:0 PM~A9 (A/C) OFF

8. Media Viewing ExampleWatching Events
Date Day Mood Start End Device Program name Type Comments Others Rating
020302 Su normal 1330 1600 T nba basketball sports
dallas mavericks go
team
none 5
020302 Su normal 1700 2100 t super bowl sports
gotta watch the
commercials
Dad 5
020402 m normal 1900 2000 t boston public drama hot teachers none 5
020402 m normal 2000 2100 t ally mcbeal drama funny lawyers none 4
020402 m normal 2300 100 V WWF RAW wrestling testosterone none 5
020502 t normal 2100 2200 t philly drama hot lawyers none 4
020602 w bored 1830 2200 t nba basketball sports GO MAVS none 5
020702 th tired 1900 2100 t
wwf
smackdown
wrestling its me soap none 5
020702 th tired 2100 2200 t ER drama good show none 4
020802 f excited 1900 2230 t olympics sports gotta watch none 4
020902 sa excited 1900 2230 t olympics sports gotta watch none 4
021002 su ecstatic 1500 1800 t
NBA allstar
game
sports
gotta see what
happens
none 3
012802 M normal 1900 2000 T Boston Public Drama hot chicks teaching none 5
012802 M normal 2000 2100 T Ally McBeal Drama hot chicks lawyering none 5

9. Multimedia Example
Digital Silhouettes (Predictive Networks)
Predicting web surfing behavior ($$$)
Microsoft (2002) track TV viewing preferences
140 data items for each user
Demographics (50)
Subcategories within gender, age, income,
education, occupation, and race
90 Content preferences
golf, music, yoga

10. Database Types / Data Models
Relational
OO
Hybrid (Object-Relational)
Temporal
Deductive
Others
Spatial, …

11. Example Data Representations
Relational
We all know…flat tables of atomic attributes with
foreign key relationships
OO
Complex data reps
multivalued, composite
Temporal
Relational model: add valid start, end dates to
each table (versions of info and when valid)
Includes time, events, durations…

12. Operations
DDL/DML (data def/manip languages)
SQL
OQL
Update operations
Built-in insert, delete, update
Stored procedures for triggers, active
(ECA) rules

13. Example Operations for
Temporal Databases
INCLUDES
Rows valid in a certain time period
BEFORE/AFTER a time condition
Set operations
Union, intersection of 2 time periods

14. Active DB
Event-Condition-Action rules
Allow for decisions to be made in the database
instead of a separate application
Relational
Implemented as triggers
Challenges
Rule consistency
(2+ rules do not contradict)
Guaranteed termination
Trigger loops (T1 <->T2)

15. Smart Home Active DB Example
Java, Postgres, Jess rules
Event classification (local&composite)
Data Manipulation Events
TV show being viewed (channel, time, genre…)
Temporal Events (instance,recurring)
Set temp to 70 degrees at 7:00am workdays
Exception Events
Power failure
Behavioral Events
Time children home from school; dinner time

16. Active DB Example (TCU)
Title Event Condition Action
TV View
Menu
TV turned on Molly is holding
remote
Display shows
matching Molly’s
preferences
Entry
Lighting
Inhabitant enters
house
Light level
<threshold
Adjust lighting to
predetermined level
Aroma-
therapy
Every Friday
night when
Hanna sits on
sofa
Always Release aroma
Night IdleJohn on sofa idle
> 15 minutes,
TV&lights are on
No other
inhabitant in
room
Turn off all devices in
the room

17. Distributed vs. Centralized
Centralized database can produce a
bottleneck
Large volume of data input
Large database
Large volume of queries
In distributed databases, data consistency,
replication, and retrieval can be more
problematic
Consistency of schemas
Retrieval in case the data location is not known
Communication overhead to ensure database
consistency

18. SmartHome
Database Architecture
Centralized vs. distributed?
Answer: Both
Central storage of high demand, persistent data
Distributed storage of low demand, dynamic data
Distributed queries
Push processing toward sensors
Adaptive, hierarchical organization
End-effector autonomy (“smart sensor”)

19. Database Systems
Commercial
DB2
Empress
Informix
Oracle
MS Access
MS SQL
Sybase
Free
Berkeley DB
PostgreSQL
MySQL

20. UTA MavHome DB
Active
Reactive & proactive (e.g., to predict)
Distributed
Information collection agents
Rules
Local Agent: what data they need to collect
Distributed: coordinate overall monitoring of collected
information
Continuous monitoring of events
Extension of SNOOP

21. Microsoft Easy Living DB
(2002)
Relational
Fast & robust, but awkward for some data
World Model DB Describes:
Computing devices
People and their personal preferences/settings
Services
Rooms and doorways
Serves as Abstraction Layer between sensors and
application that use data from sensors
e.g. new sensors no change to applications

22. Stanford Interactive
Workspace
Uses LORE
A semi-structured XML DB system
Still available, but work stopped in 2000
Data stored is catalog of (index to)
documents, images, 3-D models, application-
specific domain models

23. Sensor Database Systems
COUGAR project
www.cs.cornell.edu/database/cougar
Query processing over ad-hoc sensor
networks
Small database component (QueryProxy) at
each sensor
Sensor clusters provide local aggregations
(e.g., min, max, mean)
Assumes centralized index of all data sources

24. Siemens Netabase
“The network is the database.”
Navas and Wynblatt, ACM SIGMOD 2001
Sensor networks
Large number of data sources (105)
Volatile data and data organization
“Thin” data servers on scaled-down hardware
Netabase approach
Query decomposition
Characteristic routing (ala IP routing)
Local joins
Query evaluation

25. Siemens Netabase
www.netabasesoftware.com

26. Data Warehouses
Repositories for data mining activities
Aggregates/summaries of data help efficiency
Optimized for decision-support, not
transaction processing
Definition (Elmasri, page 900)
A subject-oriented, integrated, non-volatile, time-
variant collection of data in support of
management’s decisions”
Replace “management”, with “smart home agents”

27. Warehouse Properties
Very large: 100gigabytes to many terabytes
Tends to include historical data
Workload: mostly complex queries that access lots of data, and
do many scans, joins, aggregations.Tend to look for "the big
picture".
Updates pumped to warehouse in batches (overnight)
Data may be heavily summarized and/or consolidated in
advance (must be done in batches too, must finish overnight).
Research work has been done (e.g. "materialized views") --a small
piece of the problem.
02.15.04 from http://redbook.cs.berkeley.edu/lec28.html

28. Data Warehouses
Data Cleaning
Data Migration: simple transformation rules (replace "gender" with "sex")
Data Scrubbing: use domain-specific knowledge (e.g. zip codes) to modify
data. Try parsing and fuzzy matching from multiple sources.
Data Auditing: discover rules and relationships (or signal violations thereof).
Not unlike data mining.
Data Loading
can take a very long time! (Sorting, indexing, summarization, integrity
constraint checking, etc.) Parallelism a must.
Full load: like one big xact –change from old data to new is atomic.
Incremental loading ("refresh") makes sense for big warehouses, but
transaction model is more complex –have to break the load into lots of
transactions, and commit them periodically to avoid locking
everything.Need to be careful to keep metadata & indices consistent along
the way.
02.15.04 from http://redbook.cs.berkeley.edu/lec28.html

29. Data Warehouses
02.15.04 from http://redbook.cs.berkeley.edu/lec28.html

30. Data Mining Definition
Discovery of new information in terms of patterns or
rules from vast amounts of data
Extracts patterns that can’t readily be found by
asking the right questions (queries)
TOO MUCH DATA FOR HUMANS
Emerged from
Artificial Intelligence:Machine learning, Neural nets, Genetic
Algorithms
Statistics
Operations Research

31. Data Mining Steps
Data selection --pick the data needed
Data cleansing
Fix bad data (e.g., spelling, zip codes)
Hard to deal with missing, erroneous, conflicting, redundant
data
Enrichment
Add data (e.g., age, gender, income)
Data transformation
Aggregate (e.g., zip codes regions)
Data mining
Reporting on discovered Knowledge

32. Types of Results
Association rules
Buy diapers buy lots of beer
Sequential patterns
Buy house buy furniture within months
Classification trees
Types of buyers (upscale,bargain-conscience, …)
Why do it?
Make more money
Science & medicine

33. Data Mining Goals
Find patterns to predict future events
Find major groupings
Groupings of buyers, stars, diseases …
Find which group something belongs to
creditworthiness

34. Data Mining Results
Association rules
Classification hierarchies
Clustering
Sequential patterns
Patterns within time series
Type of result, inputs & algorithms vary
Often interested in some combination of
these types of Knowledge

35. Clustering
Unsupervised learning techniques
Training samples are unclassified
Vs. supervised learning (classification)
Drug categories for depression
Categories of TV viewers
Categories of buyers (likely, unlikely)
Categories of households?
Single male, mother/children, conventional
(M/D/kids), DINKs.

36. Sequential Patterns
Detecting associations among events
with certain temporal relationships
Example:
Cardiac bypass for blocked arteries
AND within 18 months, high blood urea
THEN kidney failure likely in next 18
months
Particularly important in smart homes

37. Sequential Pattern Discovery
Sequence of itemsets
Grocery store purchases by 1 person
(3 itemsets)
{soy milk, bread, chocolate}, {bananas,
chocolate}, {lettuce, tomato, chocolate}
2 Subsequences
{soy milk, bread, chocolate}, {bananas, chocolate},
{bananas, chocolate}, {lettuce, tomato, chocolate}

38. Sequential Pattern Discovery
The supportfor a sequence S is the % of the given
set U of sequences of which S is a subsequence.
That is: how many times does S show up?
Find all subsequences from the given sequence sets
that have a user-defined minimum support.
The sequence S1, S2, … Sn, is a predictor of “fact”
that a customer that buys itemset S1 is likely to buy
itemset S2, then S3, …
Prediction support based on frequency of this
sequence in the past
Many research issues to create good algos

39. Patterns Within Time Series
Finding 2 patterns that occur over time
2003 stock prices of Choice Homes and
Home Depot
2 products show same sales pattern in
summer but different one in winter
Solar magnetic wind patterns may predict
earth atmospheric changes

40. Time Series Pattern Discovery
Time series are sequences of events
Event could be a transaction (closing daily
stock price)
Look at sequences over n days, or
Longest period in which change is no
greater than 1%
Comparing
Must define similarity measures

41. Other Approaches in Data Mining
Neural nets
Infer a function from a set of examples
Non-parametric curve-fitting
Interpolates to solve new problems
Supervised & unsupervised algorithms
Capabilities
classification
time-series prediction
Disadvantages
can’t see what it learned (not declarative)

42. Other Approaches in Data Mining
Genetic algorithms
Set up
Representation (strings over an alphabet)
Evaluation (fitness) function
Parameters: # of generations, cross-over rate,
mutation rate, etc.
Randomized (probabilistic operators),
parallel search over search space
Used for problem solving and clustering