DATA FUNDAMENTALS

data-action-lab.com
data-action-lab.com
“Reports that say that something hasn't happened are always interesting
to me, because as we know, there are known knowns; there are things
we know that we know. There are known unknowns; that is to say,
there are things that we now know we don't know. But there are also
unknown unknowns – there are things we do not know we don't know.”
Donald Rumsfeld, US Department of Defense News Briefing, 2002

data-action-lab.com
OUTLINE

1. Data 101 – Basic Data Concepts

2. Some Practical Definitions
3. Workflows and Pipelines – the Process of Working with Data
4. Models and Systems Thinking
5. Ethical Considerations and Best Practices

data-action-lab.com
DATA 101 – BASIC DATA CONCEPTS
DATA FUNDAMENTALS

“You can have data without information, but you cannot have
information without data.”
Daniel Keys Moran (attributed)

data-action-lab.com
MODULE LEARNING OBJECTIVES

Preliminary familiarity with the following concepts:

¡ data, attribute (property, factor, variable)
¡ predictive models, explanatory models
¡ classification, class probability estimation, clustering, association rules, time series analysis,
anomaly detection, decision tree, supervised learning, unsupervised learning

Compare and contrast: data science vs analytics (Business Intelligence).

Awareness of appropriate levels of trust in models.

data-action-lab.com
WHAT IS DATA? WHERE DOES IT COME FROM?

4,529 ‘red’ 25.782 ‘Y’

data-action-lab.com
OBJECTS AND ATTRIBUTES

Object: apple
Shape: spherical
Colour: red
Function: food
Location: fridge
Owner: Jen

Remember: a person or an object is not simply the sum of its attributes!

data-action-lab.com
FROM ATTRIBUTES TO DATASETS

Attributes are fields (or columns) in a database; objects are instances (or rows)

Objects are described by their feature vector, the collection of attributes

associated with value(s) of interest

data-action-lab.com
 [https://archive.ics.uci.edu/ml/datasets/Mushroom]

POISONOUS MUSHROOMS DATASET

Amanita muscaria
Habitat: woods
Gill Size: narrow
Odor: none
Spores: white
Classification problem: Is Amanita muscaria edible, or poisonous?

data-action-lab.com
Habitat: woods
Gill Size: narrow
Odor: none
Spores: white

Classification problem: Is Amanita

muscaria edible or poisonous?

data-action-lab.com
Habitat: woods
Gill Size: narrow
Odor: none
Spores: white

Classification problem: Is Amanita

muscaria edible or poisonous?

data-action-lab.com
Habitat: woods
Gill Size: narrow
Odor: none
Spores: white

Classification problem: Is Amanita

muscaria edible or poisonous?

data-action-lab.com
Habitat: woods
Gill Size: narrow
Odor: none
Spores: white

Classification problem: Is Amanita

muscaria edible or poisonous?

data-action-lab.com
Habitat: woods
Gill Size: narrow
Odor: none
Spores: white

Classification problem: Is Amanita

muscaria edible or poisonous?

data-action-lab.com
Habitat: woods
Gill Size: narrow
Odor: none
Spores: white

Classification problem: Is Amanita

muscaria edible or poisonous?

data-action-lab.com
DISCUSSION

Would you trust an “edible” prediction?

Where is the model coming from?

What would you need to know to trust the model?

What’s the cost of making a classification mistake, in this case?

data-action-lab.com
ASKING THE RIGHT QUESTIONS

Data science is really about asking and answering questions:

¡ Analytics: “How many clicks did this link get?”
¡ Data Science: “Based on this user’s previous purchasing history, can I predict what links
they will click on the next time they access the site?”
Data mining/science models are usually predictive (not explanatory): they show
connections, but don't reveal why these exist.

Warning: not every situation calls for data science, artificial intelligence, machine
learning, or analytics.

data-action-lab.com
DATA SCIENCE/MACHINE LEARNING/A.I. TASKS

Classification and class probability estimation: which clients are likely to be repeat
customers?
Clustering: do customers form natural groups?
Association rule discovery: what books are commonly purchased together?
Others:
profiling and behaviour description; link prediction; value estimation (how much is a client
likely to spend in a restaurant); similarity matching (which prospective clients are similar to a
company's best clients?); data reduction; influence/causal modeling, etc.

data-action-lab.com
CLASSIFICATION

Classifier: If I’m presented with an object, can I classify it into

one of several predefined categories?

Many different techniques to carry this out, but the steps are
the same:
¡ Use a training set to teach the classifier how to classify.
¡ Test/validate the classifier using new data
¡ Use the classifier to classify novel instances

Some classifiers (e.g. neural nets) are very ‘black box’. They
might be good at classifying, but you don’t know why!
data-action-lab.com
TIME SERIES ANALYSIS

A simple time series:

¡ Has two variables: time + 2nd variable
¡ The second variable is sequential

What is the pattern of behaviour of

this second variable over time?
Relative to other variables?

Can we use this information to

forecast the behaviour of the variable
in the future?
data-action-lab.com
TEMPORAL PATTERNS

The goal here is our familiar analysis goals:

¡ find patterns in the data
¡ create a (mathematical) model that captures
the essence of these patterns
The patterns can be quite complex – some
fancy analysis typically required!

The data can often be broken down into

multiple component models. There are
software libraries that can help!

data-action-lab.com
ANOMALY DETECTION

Anomaly: an unexpected, unusual, atypical or statistically

unlikely event
*
Wouldn’t it be nice to have a data analysis pipeline that
alerted you when things were out of the ordinary?

Many different analytic approaches to take!

¡ Clustering
¡ Naïve Bayes

¡ Association rules deviation

¡ Ensemble techniques

data-action-lab.com
 sub-daily time window level by developing predictive classification over several different stages, as shown in Fig. 2.
models for each time window. The process made it possible to The first stage is aimed at data preparation. Data pre-processing
discover improper building energy management by detecting and is a crucial task to prepare the time series for the load pattern
analysing anomalous reduced daily patterns. To this aim, the output analysis. At this stage, the energy consumption time series are
of the process consists of a set of rules and a graphical visualisation analysed in order to identify any missing values and/or punctual
of anomalous observations/trends that deviate from the frequent/ outliers that have to be removed. The second stage of the analysis is
expected energy consumption patterns. In order to demonstrate aimed at transforming the energy consumption time series by
the effectiveness of the methodology, the whole process was tested implementing an enhanced SAX process. In detail, two preliminary
on two different buildings. hypotheses are formulated in different ways from the classic SAX
The process was conceived to be general for different types of implementation presented in Section 2. The first hypothesis is
buildings and to be useful in the post-occupancy phase so as to related to the length of the non-overlapping W windows on the

ANOMALY DETECTION CASE STUDY

support stakeholders in identifying typical building behaviour, time axis. In the literature, the time windows are generally
contextualised with respect to the boundary conditions, and assumed to have the same length [21,22,48]. However, this hy-
consequently in defining suitable energy management strategies. pothesis could cause a significant loss of information in many ap-
The implementation of this process in building automation systems plications, in terms of approximation error of the original time
could help energy managers automatically detect anomalous series. For example, when analysing building energy consumption
trends in quasi real-time in order to reduce energy waste and, data, the symbolic sub-strings have a daily length (i.e., T ¼ 24 h),
Energy 157 (2018) 336e352
consequently, the building operating costs during operation. which constrains the length that each time window can assume if

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Automated load pattern learning and anomaly detection for

enhancing energy management in smart buildings
Alfonso Capozzoli*, Marco Savino Piscitelli, Silvio Brandi, Daniele Grassi,
Gianfranco Chicco
Dipartimento Energia “Galileo Ferraris”, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129, Torino, Italy

a r t i c l e i n f o a b s t r a c t

Article history: The energy management of buildings currently offers a powerful opportunity to enhance energy effi-
Received 5 February 2018 ciency and reduce the mismatch between the actual and expected energy demand, which is often due to
Accepted 19 May 2018 an anomalous operation of the equipment and control systems. In this context, the characterisation of
Available online 21 May 2018
energy consumption patterns over time is of fundamental importance. This paper proposes a novel
methodology for the characterisation of energy time series in buildings and the identification of infre-
Keywords:
quent and unexpected energy patterns. The process is based on an enhanced Symbolic Aggregate
Energy consumption
approXimation (SAX) process, and it includes an optimised tuning of the time window width and of the
Building energy management
Adaptive symbolic aggregate approximation
symbol intervals according to the building energy behaviour. The methodology has been tested on the
Anomaly detection whole electrical load of buildings for two case studies, and its flexibility and robustness have been
Data mining confirmed. In order to demonstrate the implications for a preliminary diagnosis, some unexpected trends
Smart buildings of the total electrical load have also been discussed in a post-mining phase, using additional datasets
related to heating and cooling electrical energy needs.
The process can be used to support stakeholders in characterising building behaviour, to define
appropriate energy management strategies, and to send timely alerts based on anomaly detection
outcomes.
© 2018 Elsevier Ltd. All rights reserved. Fig. 2. e Framework for advanced energy consumption characterisation in buildings and anomalous pattern detection.

1. Introduction (i) collect a large volume of hourly or sub-hourly energy per-

formance data together with the related driving factors (e.g.,
data-action-lab.com
The increasing penetration of information and communica- internal and external conditions, occupancy, etc.) [2,3],
UNSUPERVISED LEARNING TECHNIQUES

Automated behaviours vs intelligent behaviours

Supervised: we give you some examples, you

learn from them

Unsupervised: you learn on your own, based on

what you experience

Unsupervised techniques:
¡ Association rules
¡ Recommender engines
¡ Novel categories (clustering)
data-action-lab.com
SOME PRACTICAL DEFINITIONS
DATA FUNDAMENTALS

“What’s in a name? That which we call a rose

By any other name would smell as sweet.”
W. Shakespeare, Romeo and Juliet, Act II, Scene 2

data-action-lab.com
MODULE LEARNING OBJECTIVES

Preliminary familiarity with the following concepts:

¡ data analysis
¡ data science
¡ machine learning
¡ patterns
¡ system
¡ artificial intelligence
¡ augmented intelligence

data-action-lab.com
WHAT IS DATA ANALYSIS?

Finding patterns in data

Using data to do something (answer a question, help

decision-making, predict the future, draw a conclusion)

Creating models of your data

Describing or explaining your situation (your system)

(Testing (scientific) hypotheses?)

(Carrying out calculations on data?) The more complicated the pattern, the
more complicated the analysis (?)
data-action-lab.com
WHAT IS DATA SCIENCE?

Data science is the collection of processes by which we extract useful and

actionable insights from data.
T. Kwartler (paraphrased)

Data science is the working intersection of statistics, engineering, computer

science, domain expertise, and “hacking.” It involves two main thrusts: analytics
(counting things) and inventing new techniques to draw insights from data.
H. Mason (paraphrased)

data-action-lab.com
WHAT IS MACHINE LEARNING?

Starting around the 1940s researchers began in

earnest to teach machines how to learn

The goal of machine learning was to create

machines that could learn and adapt and
respond to novel situations

A wide variety of techniques, accompanied by a

great deal of theoretical underpinning, was
created in an effort to achieve this goal

data-action-lab.com
WHAT IS ARTIFICIAL/AUGMENTED INTELLIGENCE?

Artificial Intelligence (A.I.) is non-human intelligence that has

been engineered rather than one that has evolved naturally.

Artificial intelligence research is research carried out in

pursuit of this goal.

Pragmatically speaking, A.I. is “computers carrying out tasks

that only humans can usually do”.

Augmented Intelligence is human intelligence that is

supported or enhanced by machine intelligence.
data-action-lab.com
WORKFLOWS AND PIPELINES
DATA FUNDAMENTALS

“All models are wrong. Some models are useful.”

George Box

data-action-lab.com
MODULE LEARNING OBJECTIVES

Preliminary familiarity with the following concepts:

¡ workflow and components (data collection, data exploration, etc.)
¡ analytical model
¡ data mining
¡ analytic decay
¡ data science ecosystem
¡ data science teams

Awareness of the non-linearity of the data analytical process.

data-action-lab.com
THE DATA SCIENCE “WORKFLOW”

Utilization and
Objective/ Data Data
Decision
Rationale Collection Exploration
Support

Infrastructure
Data Modeling and
and Data Communication
Preparation Analysis
Management

data-action-lab.com
THE DATA SCIENCE “WORKFLOW”

Utilization and
Objective/ Data Data
Decision
Rationale Collection Exploration
Support

Infrastructure
Data Modeling and
and Data Communication
Preparation Analysis
Management

data-action-lab.com
THE DATA ANALYSIS PROCESS

A large number of analytical models have to be generated before a final selection

can be made.

Iterative process: feature selection and data reduction may require numerous visits
to domain experts before models start yielding promising results.

Domain-specific knowledge has to be integrated in the models in order to beat

random classifiers and clustering schemes, on average.

data-action-lab.com
 [James Taylor]

CROSS INDUSTRY STANDARD PROCESS, DATA MINING

data-action-lab.com
 [James Taylor]

CROSS INDUSTRY STANDARD PROCESS, DATA MINING

data-action-lab.com
LIFE AFTER ANALYSIS

When an analysis or model is ‘released into the wild’, it can take on a life of its own.

Analysts may eventually have to relinquish control over dissemination. Results may
be misappropriated, misunderstood, or shelved. What can be done to prevent this?

Because of analytic decay, better to see the last analytical step NOT as a static
dead end, but rather as an invitation to return to the beginning of the process.

data-action-lab.com
DATA SCIENCE ECOSYSTEM

Data analysis is a team sport, with team members needing a good understanding of
both data and context
§ data management
§ data preparation
§ analysis
§ communications

Even slight improvements over a current approach can find a useful place in an
organization – data science is not solely about Big Data and disruption!

data-action-lab.com
MODELS AND SYSTEMS THINKING
DATA FUNDAMENTALS

“What if the only valid model of the Universe is the Universe itself?”
Unknown

data-action-lab.com
MODULE LEARNING OBJECTIVES

Preliminary familiarity with the following concepts:

¡ representation
¡ systems
¡ models
¡ properties
¡ knowledge gap
¡ conceptual model

data-action-lab.com
REPRESENTATION

A representation is an object that stands in for

another object.

A representation may or may not physically

resemble the object it represents.

Representations of the world help us to

understand, navigate and manipulate the world.

data-action-lab.com
 detail rigour

DATA*

DATA Data processing

purpose/goal
and reduction:
visualization
summarization
part of the data fitting
world we’re etc.
studying

explicit conceptual model + INSIGHTS

[NEW KNOWLEDGE]
(if implicit, then there can
be lack of consistency: no Automation
shared understanding) &
Implementation
data
world
collection sensor

ACTION
(based on goal)

eyeball data-action-lab.com
THINKING IN SYSTEMS TERMS

In order to understand how various aspects of the World interact with one another,
we need to carve out chunks corresponding to the aspects and define their
boundaries.

Working with other intelligences requires shared understanding of what is being

studied.

A system is made up of objects with properties that potentially change over time.
Within the system we perceive actions and evolving properties leading us to think in
terms of processes.

data-action-lab.com
THINKING IN SYSTEMS TERMS

Objects themselves have various properties. Natural processes generate (or

destroy) objects, and may change the properties of these objects over time.

We observe, quantify, and record particular values of these properties at particular

points in time.

This generates data points, capturing the underlying reality to some degree of
accuracy and error (biased or unbiased).

data-action-lab.com
IDENTIFYING GAPS IN KNOWLEDGE

A gap in knowledge is identified when we realize that what we thought we knew

about a system proves incomplete (or false).

This might happen repeatedly, at any moment in the process:

¡ data cleaning
¡ data consolidation
¡ data analysis

The solution is to be flexible. When faced with such a gap, go back, ask questions,
and modify the system representation.
data-action-lab.com
CONCEPTUAL MODELS

Exercise:
¡ assume that an acquaintance has just set foot in your living space for the first time.
¡ you are on the phone with them but not currently at home.
¡ explain to them how to go about preparing a cup of sugar.

Conceptual models are built using methodical investigation tools

¡ diagrams
¡ structured interviews
¡ structured descriptions
¡ etc.
data-action-lab.com
RELATING THE DATA TO THE SYSTEM

Is the data which has been collected and analyzed going to be of any use when it
comes to understanding the system?

This question can only be answered if we understand:

¡ how the data is collected
¡ the approximate nature of both data and system
¡ what the data represents (observations and features)

Is the combination of system and data sufficient to understand the aspects of the
world under consideration?
data-action-lab.com
TAKE-AWAYS

Certain aspects of the Universe can be approximated with the help of systems.

System models provide the basis under which data is identified and collected, but
data itself is approximate and selective.

Knowledge gaps happen. Be prepared and ready to re-visit your set-up regularly.

We often only rely on implicit conceptual modeling, but there’s danger that way.

If the data, the system, and the world are out of alignment, insights might prove
useless.
data-action-lab.com
data-action-lab.com
ETHICAL CONSIDERATIONS AND BEST PRACTICES
DATA FUNDAMENTALS

“We have flown the air like birds and swum the sea like fishes, but
have yet to learn the simple act of walking the Earth like brothers.”
Martin Luther King, Jr.

data-action-lab.com
MODULE LEARNING OBJECTIVES

Preliminary familiarity with the following concepts:

¡ ethics and best practices
¡ First Nations principles (OCAP)
¡ “do no harm“
¡ informed consent
¡ privacy
¡ model validity

data-action-lab.com
DISCUSSION

What harm can come from data?

data-action-lab.com
THE NEED FOR ETHICS

Formerly: “Wild West” mentality to data collection (and use). Whatever wasn’t
technologically forbidden was allowed.

Now: professional codes of conduct are being devised for data scientists (outline
responsible ways to practice data science).

Additional responsibility for data scientists; but also protection against being hired
to carry out questionable analyses.

Does your organization have a code of ethics for its data scientists? For its
employees?
data-action-lab.com
WHAT ARE ETHICS?

Broadly speaking, ethics refers to the study and definition of right and wrong
conducts:
§ “not […] social convention, religious beliefs, or laws”. (R.W. Paul, L. Elder)

Influential Western ethical theories:

¡ Kant's golden rule (do onto others…), consequentialism (the ends justify the means),
utilitarianism (act in order to maximize positive effect), etc.

Influential Eastern ethical theories:

¡ Confucianism, Taoism, Buddhism (?), etc.
data-action-lab.com
 [OCAP® is a registered trademark of the First Nations Information Governance Centre (FNIGC), fnigc.ca/OCAP]

WHAT ARE ETHICS?

First Nations Principles of OCAP®:

¡ Ownership
cultural knowledge, data, and information is owned by First Nations communities
¡ Control
First Nations communities have the right to control all aspects of research and information
management that impact them
¡ Access
First Nations communities must have access to information and data about themselves no
matter where it is held
¡ Possession
First Nations communities must have physical control of relevant data
data-action-lab.com
 [University of Virginia, Centre for Big Data Ethics, Law, and Policy]

ETHICS IN THE DATA CONTEXT

Data ethics questions:

¡ Who, if anyone, owns data?
¡ Are there limits to how data can be used?
¡ Are there value-biases built into certain analytics?
¡ Are there categories that should not be used in analyzing personal data?
¡ Should some data be publicly available to all researchers?

Analytically, the general is preferred to the anecdotal – decisions made on the basis
of machine learning and A.I. (security, financial, marketing, etc.) may affect real
beings in unpredictable ways.
data-action-lab.com
BEST PRACTICES

“Do No Harm”: data collected from an individual should not be used to harm the
individual.

Informed Consent:
¡ Individuals must agree to the collection and use of their data
¡ Individuals must have a real understanding of what they are consenting to, and of
possible consequences for them and others

Respect “Privacy”: excessively hard to maintain in the age of constant trawling of

the Internet for personal data.

data-action-lab.com
BEST PRACTICES

Keep Data Public: data should be kept public (all? most? any?).
Opt-In/Opt-Out: Informed consent requires the ability to opt out.
Anonymize Data: removal of id fields from data prior to analysis.
“Let the Data Speak”:
¡ no cherry picking
¡ importance of validation (more on this later)
¡ correlation and causation (more on this later, too)
¡ repeatability
data-action-lab.com
MODEL ASSESSMENT AND VALIDITY

Models should be current, useful, and valid.

Data can be used in conjunction with existing models to come to some conclusions,
or can be used to update the model itself.

At what point does one determine that the current data model is out-of-date or is
not useful anymore?

Past successes can lead to reluctance to re-assess and re-evaluate a model.

data-action-lab.com
READINGS AND REFERENCES
DATA FUNDAMENTALS

data-action-lab.com
REFERENCES

First Nations – OCAP

Wikipedia article on Semi-Supervised Learning
Wikipedia article on Supervised Learning
Wikipedia article on Reinforcement Learning
Wikipedia article on Unsupervised Learning
J. Blitzen [2017], What is it like to design a data science class?, answer on Quora
J. Taylor [2017], 4 Problems with CRISP-DM, KDNuggets.
Brin, D. [1998], The Transparent Society: Will Technology Force Us to Choose Between Privacy and
Freedom?, Perseus.

data-action-lab.com
REFERENCES

Mayer-Schönberger, V. and Cukier, K. [2013], Big Data: A Revolution That Will Transform How We Live,
Work, and Think, Eamon Dolan/Houghton Mifflin Harcourt.
Mayer-Schönberger, V. [2009], Delete: The Virtue of Forgetting in the Digital Age, Princeton University
Press.
Data Science Association, Data Science Code of Professional Conduct.
Chen, M. [2013], Is ‘Big Data’ Actually Reinforcing Social Inequalities?, The Nation.
Shin, L. [2013], How the New Field of Data Science is Grappling With Ethics, SmartPlanet.
Schutt, R. and O'Neill, C. [2013], Doing Data Science: Straight Talk From the Front Line, O'Reilly.
O'Neill, C. [2016], Weapons of Math Destruction: How Big Data Increases Inequality and Threatens
Democracy, Crown.

data-action-lab.com
REFERENCES

Chang, R.M., Kauffman, R.J., Kwon, Y. [2014], Understanding the paradigm shift to computational social
science in the presence of big data, Decision Support Systems, 63:67–80, Elsevier.
Hurlburt, G.F., Voas, J. [2014], Big Data, Networked Worlds, IEEE Computer Society.
Introna, L.D. [2007], Maintaining the reversibility of foldings: Making the ethics (politics) of information
technology visible, Ethics and Information Technology, 9:11–25, Springer.
Floridi, L. [2011], The philosophy of information, Oxford University Press.
Floridi, L. (ed) [2006], The Cambridge handbook of information and computer ethics, Cambridge
University Press, 2006.
Big Data & Ethics
Mason, H. [2012], What is a Data Scientist?, Forbes.

data-action-lab.com
REFERENCES

Schlimmer, J.S. [1987], Concept Acquisition Through Representational Adjustment (Technical Report
87-19). Department of Information and Computer Science, UCalifornia, Irvine.
Iba, W., Wogulis, J., Langley, P. [1988], Trading off Simplicity and Coverage in Incremental Concept
Learning, in Proceedings of the 5th International Conference on Machine Learning, 73-79. Ann Arbor,
Michigan: Morgan Kaufmann.
Gorelik, B. [2017], Don’t study data science as a career move; you’ll waste your time!, gorelik.net
J. Leskovec, A. Rajaraman, J. Ullman [2015] Mining of Massive Datasets, Cambridge University Press.
Hastie, T., Tibshirani, R., and J. Friedman [2008], The Elements of Statistical Learning: Data Mining,
Inference, and Prediction, 2nd ed., Springer.
Provost, F., Fawcett, T. [2013], Data Science for Business, O'Reilly.

data-action-lab.com