The University of Sydney Page 1
COMP5310: Principles of
Data Science
W1: Introduction
Presented by
Dr Ali Anaissi
School of Information Technologies
The University of Sydney Page 2
Curriculum at a glance
Whirlwind tour of:
– Data Exploration
– Data Engineering
– Data Mining & Machine Learning
– Making Decisions from Data
Focus on key activities of a data scientist
The University of Sydney Page 3
Perspectives and communication
Diverse cohort in this unit with:
– Honours degrees in non-quantitative disciplines
– Bachelors degrees in quantitative disciplines or IT
– Years of experience in industry
Doing data science requires
– Understanding application domain
– Learning, collaborating, communicating
– Product thinking
Chance to build key soft skills as well as technical skills
The University of Sydney Page 4
Questions and suggestions
We are very excited to be teaching this for the fifth year
Thank you for joining us!
Please feel free to:
– Ask questions (we should know the answer or someone who does)
– Share thoughts and suggestions on how we can improve
Questions about the MDS degree program or enrolments?
– Keiko Narushima (MDS admin officer), SIT Building, room 2E-229
– phone: 02 8627 0872 email: [email protected]
The University of Sydney Page 5
UNIT ARRANGEMENTS
The University of Sydney Page 6
Introducing Team
Lecturer
Dr Ali Anaissi
Unit Coordinator
Dr Ali Anaissi
SIT Building J12, Level 2
[email protected]
Tutors Seid Miad Zandavi
Omid Tavallaie
Hossein Moeinzadeh
The University of Sydney Page 7
Resources
Google Sheets for spreadsheet exercises [week 2]
– Please create a Google account if you don’t already have one!
Jupyter Hub accounts for Python/SQL exercises
– We will provide account details in week 3
– But we recommend you download Anaconda and PostgreSQL database
on your PC
The University of Sydney Page 8
Textbooks and readings
Data Science from Scratch. Grus. O’Reilly Media. 2015.
– Available electronically through library.
Doing Data Science. O’Neill and Schutt. O’Reilly Media. 2015.
– Available electronically through library.
The University of Sydney Page 9
Learn Python and SQL with Grok
– Exercises will use Python from week 3
– We provide self-guided Python learning through Grok
– Grok learning modules are available now in Canvas under
Assignments folder
– Please complete (sooner is better, week 5 at latest)
The University of Sydney Page 10
Find everything on Canvas
– The web site for this unit is on Canvas
– Use it to access contacts, schedule, readings, slides, etc
– Participate in Q&A with instructors and classmates
https://canvas.sydney.edu.au
The University of Sydney Page 11
ASSESSMENTS
The University of Sydney Page 12
Assessment
– 10%: Participation
– 10%: Project stage 1
– 15%: Project stage 2
– 5%: Project stage 3
– 60%: Final exam
The University of Sydney Page 13
Participation
Objective
Ensure everybody is keeping up.
Requirements
Submit code at end of each exercise
Complete Grok exercises (not marked)
Output
Code/spreadsheets from exercises
Marking
10% of overall mark
The University of Sydney Page 14
Project stage 1: Explore, Clean, Pitch
Objective
Explore a data set and define a research
question based on research/business
requirement.
Activities
Choose a data set
Explore, summarise and prepare data
Define problem, specify requirements
Output
2-page report summarizing problem
analysis and proposal (plus code)
Marking
10% of overall mark (report and code)
The University of Sydney Page 15
Project stage 2 and 3: Experiment, Quantify, Report
Objective
Define an experimental framework and
complete analysis/visualisation, data
mining, machine learning, etc.
Activities
Define experimental framework
Perform analysis or build tool
Describe evaluation and conclusions
Output
4-page report describing framework,
analysis and conclusions (plus code)
Presentation (2-3/3-4 mins)
Marking
20% of overall mark
– 15% report and code
– 5% presentation
The University of Sydney Page 16
Final exam
Objective
Assess understanding of unit material,
ability to frame data problems
scientifically and critical thinking about
claims made based on data
Activities
Answer questions about lecture materials
Practical excises and SQL queries
Describe an approach to answering a
question with data
Critique a claim made based on data
Format
Written examination
Must get 40% on exam to pass unit per
SIT policy
Marking
60% of overall mark
cap on final mark which cannot exceed
exam mark by more than 10 marks
The University of Sydney Page 17
Lecture plan
– W1: Introductions and housekeeping
– W2: Data exploration (spreadsheets)
– W3: Data exploration (Python)
– W4: Cleaning and storing data
– W5: Querying and summarising data
– W6: Hypothesis testing
Project stage 1 due
– W7: Data Mining - Association Rules and
Dimensionality Reduction
– W8: Data Mining - Clustering
– W9: Machine Learning – Regression
– W10: Machine Learning – Classification
– W11: Unstructured Data
– W12: Review
Project stage 2 and 3 due
– Exam
The University of Sydney Page 18
LATENESS AND PLAGIARISM
The University of Sydney Page 19
Recipe for success
– Attend scheduled classes except for illness, emergency, etc
– Plan 6-9 hours per week for preparation, practice, project, etc
– Participate in classes and forums with respect and humility
– Submit assessments on time
– Let us know if any concerns, e.g., if you are falling behind
The University of Sydney Page 20
Special consideration (University policy)
– If your performance on assessments is affected by illness or
misadventure
– Follow proper bureaucratic procedures
– Have professional practitioner sign special USyd form
– Submit application for special consideration online, upload scans
– Note you have only a quite short deadline for applying
– http://sydney.edu.au/current_students/special_consideration/
– Notify us by email as soon as anything begins to go wrong
– There is a similar process if you need special arrangements for
religious observance, military service, representative sports, etc
The University of Sydney Page 21
Penalty for lateness
– If you have not been granted special consideration
– Penalty is 5% of awarded marks per day
– Maximum 10 days late, then 0 points
– Examples:
– Work would have scored 60% and is 1 hour late: 57%
– Work would have scored 70% and is 28 hours late: 63%
– Recommendation: submit early; submit often
The University of Sydney Page 22
Academic integrity (University policy)
“The University of Sydney is unequivocally opposed to, and
intolerant of, plagiarism and academic dishonesty.
Academic dishonesty means seeking to obtain or obtaining
academic advantage for oneself or for others (including in the
assessment or publication of work) by dishonest or unfair means.
Plagiarism means presenting another person’s work as one’s own
work by presenting, copying or reproducing it without appropriate
acknowledgement of the source.”
http://sydney.edu.au/elearning/student/EI/index.shtml
The University of Sydney Page 23
Academic integrity (University policy)
– Submitted work is compared against other work
– Turnitin for textual tasks (through eLearning)
– other systems for code
– Penalties for academic dishonesty or plagiarism can be severe
– Complete required self-education AHEM1001
The University of Sydney Page 24
INTRODUCTIONS AND
BACKGROUNDS
The University of Sydney Page 25
Exercise: Survey of skills and interests
https://goo.gl/BgVnjR
(link on Canvas)
Survey – Individual Responses
What kind of role would you like (Data Engineer/Scientist, Analyst, etc)?
What are the three most important data analytics skills?
We’ll explore this data in week 2 exercises!
The University of Sydney Page 26
WHAT IS DATA SCIENCE?
The University of Sydney Page 27
Data Scientists
build intelligent
systems to derive
knowledge
from data.
The University of Sydney Page 28
http://www.marketingdistillery.com/2014/11/
29/is-data-science-a-buzzword-modern-data-
scientist-defined/
Data Science skills
Data scientists help organisations:
– understand their data,
– ask meaningful questions,
– derive transformative insights,
– lead empirically grounded decision
making.
The University of Sydney Page 29
Cross Industry Standard Process for Data Mining
(CRISP-DM)
By Kenneth Jensen - Own work based on:
ftp://public.dhe.ibm.com/software/analytics/spss/documentation/modeler/18.0/en/ModelerCRISPDM.pdf (Figure 1), CC BY-SA
3.0, https://commons.wikimedia.org/w/index.php?curid=24930610
The University of Sydney Page 30
Business Understanding Phase
– Business objective
– Understand business processes
– Associated costs/pain
– Assess situation
– Define the success criteria
– Data science goals
– Project plan
– List assumptions and risk (technical/financial/business/ organizational)
factors
The University of Sydney Page 31
Some example goals
– Farmer wants advice on what fertilizer to use, to maximize crop
yield
– Bank wants to automatically flag some credit card purchases
as potentially fraudulent, to delay payment till checks have
been made
– Biologist wants to be able to find out which species of micro-
organism are present in a location, given a list of protein
fragments found in an environmental sample
The University of Sydney Page 32
Some example goals (cont’d)
– Doctor wants to determine whether a patient is likely to have a
particular disease, given results of tests (none of which is
perfect)
– Designer wants a car that brakes automatically when a
pedestrian steps in front
The University of Sydney Page 33
Data Understanding Phase
– Collect Data
– What are the data sources?
• Original sources (these all will contain errors!):
– sensors (measure the world)
– surveys (ask people)
– digital logs (track IT activities)
• Secondary sources
– other scholars, organizations, etc
– data may already be summarized, transformed,
cleaned, etc
The University of Sydney Page 34
Examples of datasets
– Census
– raw data has individual level demographics etc
– available summaries combine these into counts in a suburb etc
– Crop observations
– many plantings, with many features (seed type, date, weather, soil,
fertilizer etc), and resulting crop yields
– Credit card histories
– lots of transactions of many users, with many features, some transactions
were reported as fraudulent
– Medical records
– lots of patients, their test results, diagnoses
The University of Sydney Page 35
Data Understanding Phase
– Data Description
– Document data quality issues
– Compute basic statistics
– Data Exploration
– How is it structured? What is the meaning of the different features?
• eg is temperature the daily maximum, monthly average, at some
specific time? is income measured in actual dollars or inflation-adjusted
ones?
– Simple univariate data plots/distributions
– Investigate attribute interactions
• Can you find patterns connecting different features?
– Data Quality Issues
The University of Sydney Page 36
Data Preparation Phase
– Integrate Data
– Joining multiple data tables
– Summarisation/aggregation of data
– Select Data
– Attribute subset selection
• Rationale for Inclusion/Exclusion
– Data sampling
• Training/Validation and Test sets
The University of Sydney Page 37
Data Preparation Phase (cont’d)
– Data Transformation
– Using functions such as log
– Factor/Principal Components analysis
– Normalization/Discretization/Binarization
– Clean Data
– Handling missing values/Outliers
– Data Construction
– Derived Attributes
The University of Sydney Page 38
The Modelling Phase
– Select of the appropriate modelling technique
– Dependent on
• Data mining problem type
• Output requirements
– Develop a testing regime
– Sampling
• Verify samples have similar characteristics and are
representative of the population
The University of Sydney Page 39
The Modelling Phase (cont’d)
– Build Model
– Choose initial parameter settings
– Study model behaviour
• Sensitivity analysis
– Assess the model
– Beware of over-fitting
– Investigate the error distribution
• Identify segments of the state space where the model is less effective
– Iteratively adjust parameter settings
• Document reasons of these changes
The University of Sydney Page 40
Examples of Models
– Model to predict the purity of the environment based on
carbon level (Regression prediction model)
– Model to classify a person whether he is cheating in his tax
return or not (Classification prediction model).
– Model to find hidden patterns and association rules in the
basket market analysis (Clustering or association rules).
– Model to detect anomalies or outliers such as spam emails
(Classification prediction model).
The University of Sydney Page 41
The Evaluation Phase
– Validate Model
– Human evaluation of results by domain experts
– Evaluate usefulness of results from business perspective
• Define control groups
• Expected Return on Investment
– Review Process
– Determine next steps
– Potential for deployment
– Metrics for success of deployment
The University of Sydney Page 42
The Deployment Phase
– Knowledge Deployment is specific to objectives
– Knowledge Presentation
– Automated pre-processing of live data feeds
– Generation of a report
• Online/Offline
– Monitoring and evaluation of effectiveness
The University of Sydney Page 43
DATA SCIENCE PROJECTS
The University of Sydney Page 44
http://www.bloomberg.com/news/articles/2013-10-30/ups-uses-big-
data-to-make-routes-more-efficient-save-gas
Example: Reducing costs through route optimisation
– Use customer, vehicle and
delivery data
– 1 mile less per day for
every driver saves $50
million p.a. in fuel,
maintenance and time
– Less idling, e.g., by avoiding
left turns, saved 1.6 million
gallons of fuel in 2012
The University of Sydney Page 45
Example: Structural Health Monitoring (SHM)
 Time-based maintenance:
• Preventative maintenance schedules
• Too early or too late
 SHM:
• Condition-based maintenance using
sensors
• Data-driven approach establishes
model from data, using machine
learning techniques.
The University of Sydney Page 46
http://arxiv.org/pdf/1508.03965v1.pdf
Example: Preventative policing
– Given social network from
arrest records, geographic,
temporal data
– Predict whether a person is
likely to be involved in crime
– Chicago police using to issue
preemptive warnings:
“We’re watching you”
The University of Sydney Page 47
Example: Road Condition Assessment from Vehicle-
mounted Sensor
Data Acquisition
Machine Learning AnalysisFeature Extraction
Road Health Score
Excitation
The University of Sydney Page 48
WHERE DO I GET DATA?
The University of Sydney Page 49
Source Example: UCI Machine Learning Repository
Datasets
About
The UCI Machine Learning Repository is a
collection of databases, domain theories,
and data generators that are used by the
machine learning community for the
empirical analysis of machine learning
algorithms.
URL
https://archive.ics.uci.edu/ml/datasets.html
Data sets
– Classification
– Breast Cancer
– Diabetes
– Letter Recognition …etc.
– Regression
– Forest Fires
– Buzz in social media ..etc.
– Clustering
– Bag of Words
– Sponge …etc.
The University of Sydney Page 50
Source Example: Kaggle Datasets
About
Kaggle is an online platform for data
science competitions. Some data sets are
publicly available.
URL
https://www.kaggle.com/datasets
Data sets
– Amazon fine food reviews
– Health insurance marketplace
– World food facts
– Ocean ship logbooks
– Reddit comments
– Hillary Clinton’s emails
– GOP debate Twitter sentiment
– NIPS 2015 papers
The University of Sydney Page 51
Source Example: AIHW Data
About
Australian Institute of Health & Welfare
collects data that provide insight into the
health and wellbeing of the multifaceted
Australian population.
URL
http://www.aihw.gov.au/data-by-subject/
Data sets
– Alcohol, Tobacco & Drugs
– Cancer
– Children’s health
– Height & weight
– Hospitals
– Indigenous health
– Mental health
– Lots more!
The University of Sydney Page 52
Source Example: Reddit comments
About
Reddit is a social news web site that
functions like an online bulletin board.
URL
https://www.reddit.com/r/datasets/comm
ents/3bxlg7/i_have_every_publicly_avai
lable_reddit_comment
Data sets
– 1.7 billion public comments
The University of Sydney Page 53
REVIEW
The University of Sydney Page 54
W1 Review: Introductions and housekeeping
Objective
Housekeeping; Learn about backgrounds
and goals; Define data science.
Lecture
– Welcome, introductions
– Unit overview, assessment, resources
– Learning Python with Grok
– Discuss definitions/scope of data
science
Readings
– Data Science from Scratch: Ch 1
– Is being a data scientist really the
best job in America?
– 8 skills you need to be a data scientist
Exercises
– Introductions / interviews
– Interests / definitions
TODO in W1
– Grok Python modules 1-3
– Fill out & submit background survey
– Choose possible project data
The University of Sydney Page 55
Formulating a COMP5310 project (Stage 1 & 2)
– By next week:
– Identify possible problems and data sets
– Think about questions the data can answer
– Other possible data sets…
The University of Sydney Page 56
Source Example: Yahoo Webscope
About
The Yahoo Webscope program is a
reference library of data sets for non-
commercial use by academics.
URL
http://webscope.sandbox.yahoo.com/
Data sets
– 13.5 TB of user interaction data
– Search engine query logs
– Q&A forum data
– Query entity disambiguation
The University of Sydney Page 57
Source Example: GovHack Data
About
GovHack is an annual event that brings
people together to innovate with open
government data. They list many data sets
from Australia and New Zealand.
URL
http://portal.govhack.org/datasets.html
https://data.gov.au/
Data sets
– ABC news and TV archives
– Australian census data
– Labour, industry, transport data
– Health and welfare data
– Various CSIRO data sets
– Finance, IP, geoscience, archives, etc
The University of Sydney Page 58
NEXT TIME
The University of Sydney Page 59
Next week: Data exploration with spreadsheets
Objective
Use interactive tools to explore a new
data set quickly.
Lecture
– Data types, cleaning, preprocessing
– Descriptive statistics, e.g., mean,
stddev, median
– Descriptive visualisation, e.g.,
scatterplots, histograms
Readings
– Data Science from Scratch: Ch 2-3
Exercises
– Google Sheets: Visualisation
– Google Sheets: Descriptive stats
TODO for W2
– Grok Python modules 1-3
– Make sure you answered today's
background survey
– Explore project data
– GET YOUR GOOGLE ACCOUNT!
The University of Sydney Page 60
Thanks

欢迎咨询51作业君