Question 3

Techniques for policy informatics: Data mining and Data

visualization: Support your research with references (Harvard) and
examples of policies
Technological advances in various computing fields, AI included, either have already had a

dramatic impact or are perceived as potential game changers in many applications. Take big data

and social media as cases in point.

Policy Informatics for Smart Policy-Making Daniel Zeng, University of Arizona and Chinese

Academy of Sciences promise to change how the scientific enterprise operates and innovates. In

industry settings, new waves of products and services based on these technologies, many by

startups touted as tomorrow’s Google and Facebook, are entering the marketplace, improving

productivity in old industry sectors and opening up new opportunities for future businesses yet to

be defined. In the public sector, these technologies are starting to make similarly profound

impact and promise to potentially revolutionize policy-making. Compared with applications

stemming from the scientific community or private sector, public-sector applications tend to take

a more conservative approach (likely rightfully so) when adapting and adopting technology. As a

result, it’s still too early to pinpoint full-scale redesign of policy-making approaches in major

public decision-making areas, or to discuss completed success stories. Yet, bits and pieces of the

next generation of IT-driven policy-making have been emerging for some time. It’s already

common knowledge that social media provides great potential for policy makers to gauge public

opinion. Researchers and analysts routinely study social media content, such as Twitter feeds, to

characterize patterns of public sentiment on various issues with policy relevance, and in some

cases, to identify how emotion or influence propagates through online social networks.

Understanding gained through such analyses complements the traditional approach, which is
largely based on polling, and can inform policy-making. In more specific public-sector

application scenarios such as emergency response, a whole suite of information technologies,

including but not limited to sensor networks and social media, are being integrated to provide

faster and finer-granularity assessment of a situation. In addition, these technologies serve as the

enabling mechanism for coordination among people as well as resources, often in a distributed

fashion, and provide timely feedback to decision makers either during the planning phase or after

the implementation of involved policies and decisions. In public health crisis management, most

notably during the most recent Ebola epidemic, and the 2009 H1N1 (swine flu) pandemic, social

media has been heavily studied, with all kinds of simulation-based predictive models developed.

Despite the fact that most of these models performed poorly (and were often outrageously

wrong), the public health community has argued for the value of such models, from the policy

standpoint.

Policy Informatics

Policy makers have traditionally relied on intuition, experience, small-sample human contact,

polls, and media outlets to gauge society’s “pulse.” As I argued above, IT is providing innovative

alternatives to improve and complement the traditional approach.

The field of policy informatics is emerging to cross-fertilize between computational sciences and

public administration and policy, and to advance the framework of and infrastructural support for

public policy-making. Current policy informatics research streams emphasize traditional

informatics research in the context of public policy and administration. New do-main-specific

information collection and analysis approaches are being developed to meet the needs of

complex policy and administration problems. Work on social media, population-scale big open

data, and temporal-spatial-network visualization has received a lot of attention lately; more
applied research investigates technology adoption issues in governance processes. Also gaining

momentum is research on new data-driven decision making models; complex systems view of

governance; collective intelligence; behavioral studies of policy-makers, policy-making

processes, and the public; and persuasive technologies. From Informed Policy Making to Smart

Policy-Making Policy informatics is in its early stage of development, yet many of its concepts

and techniques have already met with success in the policy research community and in practice.

It’s safe to characterize the state of the art as being largely informatics and data science-based,

focusing on better situational awareness.

In this sense, policy informatics is already using technological means to enable informed policy-

making. Granted, providing principled informed policy-making frameworks and tools to policy

makers can have enormous implications; concerted efforts from multiple disciplines are still

needed in the years and decades to come, to perfect such frameworks and tools, and promote

their adoption. In the meantime, it makes a lot of sense for the research community to look

beyond informed policy-making to explore how policy informatics can help build the foundation

for even better and more advanced policy-making, something more akin to smart policy-making.

Of course, smart policy-making is built on top of informed policy-making, which promises

effective, real-time situational awareness and analysis capabilities to make use of data.

What advanced capabilities, then, will differentiate smart policy-making? To come up with a

definitive set of such differentiators won’t be possible due to the topic’s emerging nature. But

from the current literature and ongoing discussions among academics and practitioners, it isn’t

too difficult to venture on some novel aspects or even pillars of next-generation smart

policymaking:

 Informed policy-making focuses on what has happened and what is happening. Smart
 policy-making needs to take into consideration additional information, such as what

might happen down the road. In other words, smart policymaking entails a much more

proactive framework.

 In informed policy-making, data processing tends to be treated as an independent

capability, emphasizing various engineering aspects of data sharing and mining. In smart

policy-making, the integration between data and the domain is expected to be much

tighter. Various kinds of behavioral, affect, and root-cause analyses, at both the

individual and group/population levels, would need to be carried out in particular policy

contexts.

 With a significantly improved understanding of the policy environment, and much more

detailed, data-driven models, we can expect possible major changes as to the framing,

decision-making framework, and evaluation mechanism of policy alternatives. As such,

it’s conceivable that new decision-making tools, in addition to informatics and situational

awareness tools, will play a crucial role in smart policy-making. Since the development

of policy informatics, AI has been a major contributor. A significant portion of technical

work supporting informed policymaking can be classified as applied AI research. Smart

policy-making is expected to create more exciting and novel research problems for AI

researchers, challenging the start-of-the-art predictive analytics, behavioral analysis, and

multi-agent simulation, among others. In an era full of opportunities for AI researchers in

applications such as self-driving cars, robotics, Siri, and Cortana, it’s important for the AI

community not to forget policy-making as a fruitful area with great potentials.

i) Data mining techniques

Data itself is a collection of raw facts such as words, numbers, descriptions, observations, and

measurements of things since its raw data which means for it to be processed and made

meaningful some data mining techniques need to be applied. Datamining has a lot of techniques

that can be used effectively depending on the type of data set. Datamining is used to come up

with trends and patterns, for this to come to play one of the data mining techniques has to be

applied depending on the type of data available.

Tracking patterns.

One must be able to track patterns in a given data set, this is one of the most recommended skills

that a data miner should have. This is usually involves the recognition of some reputations or

trends in your data happening at regular intervals. For example, you might see that a certain route

is taken by most people during summer or NUST students board taxis in the morning than in the

evening.

Classification.

Classification is a more complex data mining technique that forces you to collect various

attributes together into discernable categories, which you can then use to draw further

conclusions, or serve some function. For example, if you’re evaluating data on individual

student’s attendance and number of courses passed, you might be able to classify them as “low,”

“moderate,” or “high” performance. You could then use these classifications to learn even more

about those Students. Association.

Association is related to coming up with patterns, but is more specific to dependently linked

variables thus one has to identify certain patterns that are linked to one another for example in a

shop might realize that a person who buys bread always buys coffee, these items highly

correlated with another thus this technique is mostly used in shelf arrangement in shops, you’ll
look for specific events or attributes that go hand in hand.

Outlier detection.

Most of the data sets are filled with outliers and in many cases, simply recognizing the

overarching pattern can’t give you a clear understanding of your data set. You also need to be

able to identify those anomalies, or outliers in your data. For example, in a class usually people

who excuse themselves during lecture hours are males but one strange week during the month,

there’s a huge spike in females who excuse themselves during the lectures, you’ll want to

investigate the spike and see what drove it, so you can either replicate it or better understand

your audience in the process.

Clustering.

Clustering is very similar to classification, but involves grouping chunks of data together based

on their similarities. For example, you might choose to cluster different items of your audience

into different packets based on how often they tend to shop at your store.
Regression.

Regression, used primarily as a form of planning and modeling, is used to identify the likelihood

of a certain variable, given the presence of other variables. For example, you could use it to

project a certain price, based on other factors like availability, consumer demand, and

competition. More specifically, regression’s main focus is to help you uncover the exact

relationship between two (or more) variables in a given data set.

Prediction.

Prediction is one of the most valuable data mining techniques, since it’s used to project the types

of data you’ll see in the future. In many cases, just recognizing and understanding historical

trends is enough to chart a somewhat accurate prediction of what will happen in the future. For

example, you might review Students payment history and past payments to predict whether

they’ll be a credit risk in the future.

iii) Data Visualization Techniques

Data visualization is the graphical representation of information, data in a data set can be

represented using different data visualization tools if you are working with massive amounts of

data, one challenge is how to display output in a way that’s not overwhelming Data visualization
can help you to create better and more appealing business reports, maximizing the potential of

your analysis. If you want the attention of your clients and colleagues, you need to learn modern

data visualization models to improve the quality of your presentation.

Best Data Visualization Tactics

Data visualization can communicate complex information in a way that is easier to interpret by

turning information into visually engaging images and stories. It enables you to highlight the

most relevant conclusions from what would otherwise be considered a huge pile of worthless

documents.

Get to know the audience

In Data visualization you need to match the audience, thus you need to know literate is your

audience. Some people cannot read a simple bar graph or pie chart thus you have to make it as

simple as possible. Even those who want to interpret data should be able to draw conclusions

from the given data. (Neely,2007 ) explained it concisely: “If you are dealing with

inexperienced clients, stay away from advanced solutions. But if you are meeting highly skilled

professionals, going beyond pies and charts is mandatory”. Therefore, you must get to know the

audience you face and give them materials they can digest successfully.

Think about the content

What you want to present is as important as who you are showing it to. There are 4 basic ways  to

approach data visualization: Relationships: Shows the connections thus mutual impact between

specific elements (such as course and Program). Scatter plot is the best choice in this case.

Timeframe: Line graphics suits perfectly if you want to show how certain phenomenon is

developing over time. Composition: This technique is developed to reveal the structure of a

single unit, showing its constitutive elements. A pie chart is the simplest way to do this but if you

want a more distinguished data visualization, go for the 100% stacked horizontal bar graph or a

slope graph. Comparisons: Bar charts are the usual suspect if you want to compare two or more

values.

Mind the colors

Although it seems irrelevant, the colors that are chosen have a strong impact the overall

effectiveness of your data visualization model. The contrasts between the opposing elements,

emphasizing the differences among these features. People mostly use red, green, blue, and

yellow because they can be recognized and distinguished easily, should not mix the colors too

much because it creates confusion among viewers and interferes with already established

patterns. For example, if you used red to mark negative trends and green to highlight positive

outcomes, don’t change the style throughout the document.

Use interactive maps

Data visualization can become a source of valuable digital content, which demands adding

interactive elements to the presentation. Interactive maps play the major role in that regard

because they allow users to engage and look only for information that they really need.
Interactive maps enable users to wander around the chart, zoom in and out, identify special

elements upon click, get a 360-degree overview, and many other interesting features. Creating

such maps is a highly complex process but it will definitely leave a great impression on your

clients or customers.

As the matter of fact, interactive maps had already become a standard technique for the vast

majority of companies and websites, with the likes of Google, Booking.com, or National

Geographic setting a good example for data visualization community.

References

Andersen, et al. (2007) Group model building: Problem structuring, policy simulation and

decision support. Journal of the Operational Research Society 58(5), p691--694.

Bryson, J. M. (2004). What to do when stakeholders matter. Public Management Review, 6(1),

p21--53.

Dawes, S. and Helbig, N. (forthcoming). The Value and Limits of Government Information

Resources for Policy Informatics in De Souza, K. and Johnston E., Policy Informatics, MIT

Press.

eGovPoliNet--The policy community Date accessed 24 September 2018, http://www.policy-

community.eu/

Janssen, M. and Klievink, B. (2010). Gaming and simulation for transforming and reengineering

government: Towards a research agenda. Transforming Government People Process and Policy

0 4, p132--137.
Koliba C. and Zia A. (forthcoming) Governance Informatics: Using Computer Simulation

Models to Deepen Situational Awareness and Governance Design Considerations in De Souza,

K. and Johnston E., Policy Informatics, MIT Press.

Pedercini, M. (2005). Potential Contributions of Existing Computer-based Models to

Comparative Assessment of Development Options, Center for Conservation and Government.

Downloaded from http://www.millennium-institute.org/resources/elibrary/papers/CI-

ModelsReport-Final.pdf April 15, 2013.