Speakers

Big data, ...everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is going it, so everyone claims they are doing it. Whether you have terabytes of data or not, a sophisticated
analysis will quickly push the limits of your hardware and your engineering team.
Indeed, data, big or small, doesn't speak for itself. Machine learning can build from data predictive models with optimal decision logic. But machine learning can also be a tricky and expensive endeavor.
In this talk, I will discuss how the combination of a powerful Python machine learning stack, with scikit-learn and joblib at it's core, and simple data-processing patterns can be used for leading-edge data analysis with efficiency in mind. The efficiency that matters is not only about using hardware well, but also about human factors: quick, agile, short-cycles to understand the data best.

In this presentation Matt speaks about how he discovered R and why he loves it. He shares his experiences using R since 2002 and also tells the story of data.table, the package he created for making manipulating data in R comfortable.

Open Data is more than a trend in Europe. It brings new business opportunities, increases trust in governments, and enables innovation. The recent amendment of the Directive 2003/98/EC on Public Sector Information re-use includes important changes, such as considering Open Data as a genuine right for citizens. July 2015, deadline to transpose the Directive into national laws, will be an important milestone for innovation, transparency and democracy in Europe.

This session will be focusing on use case of text mining twitter twits. After establishing certified and authenticated connection between R and Twitter, using power of R library for text mining, one can extract the terms, compare the documents create statistics. In this session we will also use R for doing segmentation of twits in order to see, what groups of followers an account has and what are the main topics as well as most common words relevant for each group.

dplyr is an increasingly popular R package from Hadley Wickham, that presents itself as a grammar for data manipulation. The package works with simple verbs — functions that take data frames as first argument — that each have a well identified role in the data manipulation pipeline: filtering rows, selecting columns, mutating columns, grouping by one or several variables, summarizing and joining. Even though Romain is more focused on the C++ internals of dplyr, this presentation is a quick sightseeing tour of the various dplyr verbs.

Even though analyzing unstructured text data is rapidly becoming an everyday task for data scientists, it is never a routine task due to the special challenges that free text presents for automated analysis. Fortunately, python has an ecosystem of powerful packages which make the natural language processing workflow relatively smooth and manageable.
In this session, we are going to show through examples how the numpy, pandas, pytables, scikit-learn, and nltk libraries can be used to solve some typical natural language processing problems.

The scientific community itself generates significant amount of data, which is publicly available via literature databases. By collecting these data on publications, coauthor networks can be constructed, yet this process involves data mining, cleaning and visualization in a larger scale. The flexibility of R provides a suitable framework to develop and use a script which is capable the cover the whole analysis, from data entry to visualization. The R-script is presented via a working example of the Hungarian health economist community.

“But SNA is not just a methodology; it is a unique perspective on how society functions. Instead of focusing on individuals and their attributes, or on macroscopic social structures, it centers on relations between individuals, groups, or social institutions.” The aim of the presentation is to give a basic understanding about the social networks, it's representations, analysis and about the meanings of the derived metrics. The R programming languages and it's relevant packages will be used as one of the most widely used open source data analysis platform.

Collecting and harmonising data from various formats and sources, reproducible and flexible generation of graphics for analysis and communication are our core statistical activities. Re-occurring tasks suggest the establishment of reporting systems to better control speed and quality of outputs. R user community developments entail very suitable tools to rapidly set up prototype systems with sufficient performance. After definition of in- house usage standards for open sources tools, duplication of efforts can be reduced when researchers document their procedures and make them available using version control and code sharing platforms.

Forex market (short introduction) Introduction into the FOREX market. How it works, how you can be part of it. The reason choosing Forex. Why „R” The reason behind using „R” for this project. Showing the environment and packages applied in development. Trading idea. Throwing up a hypothetic strategy idea, after some analysis. First test of the strategy. Demonstration of a trading simulator written by ourself applied to the strategy detailed above. Optimalization (generate_heatmap). As the first simulation showes some hope we try to see if there is an optimal parameter set for the strategy. Backtest in large. „Large scale backtesting” using the optimal parameter set derived above for a long dataset. Problems and solutions. Backtest with PBO. There is a great problem developing automated strategies: Overfitting. We try to solve this by using the package PBO. Optimalization + forward testing. Now we are convinced that our strategy worth testing for real-life usage. Showing the forward testing stage here. Real life example (MT4 + „R”) Putting the whole into work. Connecting together Metatrader and „R” showing it working real time.

Traditionally BI focused on data owned by the organization, data within the firewall. However, in the last 15 years there’s been an explosion of information outside the firewall. Managing that data can be helpful for recruiting, marketing, sales development, but also for product development, by giving companies better information about what potential customers want.
Classical BI uses technologies, processes, and applications to analyze mostly internal, structured data and business processes. However, data on the Internet comes in various formats: mostly unstructured or semi-structured, often noisy or not completely reliable. The density of information is lower.
This presentation provides an overview of these challenges and how the combination of natural language processing, machine learning and data analytics can be used to overcome them for collecting valuable pieces of information from this vast ocean of data.

When the Enron energy corporation went bankrupt, university researchers acquired a subset of their emails. Felienne Hermans of Delft University has examined those emails, specifically looking for attached spreadsheets.
Felienne found that emailing spreadsheets was very common within Enron, because the set of emails spanning 15 months of emails contained over 50.000 attached spreadsheets! In this session Felienne will explain:

• The most interesting results from the Enron set
• What conclusions we can draw from it
• Methods and techniques to battle spreadsheet problems

The growing amounts of data and the desire to gain insights from them, demands for easy-to-use tools to perform the tasks of exploration, evaluation and communication of information. As a key component of such "macroscope" interactive, dynamic visualizations can help the user to gain an intuitively accessible, memorable and actionable understanding of the information. But, if you’re now thinking about the horrors of complicated business intelligence applications of the past or futuristic holographic interfaces à la Iron Man, think again. Today’s consumer applications are presenting vast amounts of information that the user needs to handle with ease and confidence.

In the process of creating those kinds of applications we see two different approaches collide: On the one hand, the representation of data requires an intensive examination of the structure and texture of the underlying data set. On the other hand, a user-centered design process must be followed to ensure usefulness and usability for the human using it. Based on case studies from our daily work, I will talk about the pitfalls and highlights of marrying these two principles into a design process that is as robust as possible yet as flexible as necessary.

How do you get your company to move away from classical, “heavy” BI platforms? How do you motivate people to use more user-oriented BI tools and approach data analysis in a smarter way? What issues are you likely to encounter with business and IT teams, and how can you overcome those hurdles?

Join this session to learn about strategies to encourage the adoption of new BI tools in your organization, which include developing simple solutions to tedious tasks and showing your team that BI software can be more than just an application for answering serious business questions. Yes, data analysis and visualization can be fun!

The talk will feature a case study from Vodafone, where a small team of enthusiastic data-minded people were able to achieve great performance using Tableau Software’s products. You will see a live demo of interactive dashboards used in project management, understand how storytelling with data can help your business, and smile at examples of fun BI.

Visualizations are not created equally and different situations require different solutions. In this talk Jan Willem will present a practical model to categorize different types of visualizations and how
design decisions differ in each situation, with an extra emphasis on visualizations in a business situation.

Version control and continuous integration have become de-facto standards in software engineering. Source control gives software engineers control over changes to their source code, removes the friction of team collaboration and is vitally important for locating and fixing bugs. For developers, it is unthinkable to work without revision control. Yet when it comes to spreadsheets, the best we seem to have come up with so far is file name timestamping and keeping long lists of file versions.

In this talk I will give an overview of why and how to think about version control for spreadsheets - with a strong emphasis on non developer folks. I will also discuss technical and business lessons learned from building a product and business in a market that is considered stagnant by many people.

I have seen not once in my life genious researches and professional analyses swept under the rug, perfectly set up tests and their results ignored, action plans which were the exact opposites of survey results being deployed and failed. Without exception these were the consequences of wrongly communicated data within the organizations. Communication of data can happen on several levels and each level can be described with its own typical mistake - starting with the definition of wrong key messages or key results through using too simple or too complicated charts to choosing (or not choosing) the right communication channel (email, presentation, etc.).
I truly believe that there are no such things as dumb decision makers or short sighted bosses.... only poorly communicated data. If we manage to avoid these classical mistakes highlighted in this presentation, successful decisions will be made far more easily and everyone will be happy. I will use entertaining but instructive examples brought from the world of startups, multinational and small-middle sized companies.

In this presentation the speaker will present different methods and solutions that they have developed during the last 2 years to provide a synthetic visualization of large amount textual data (in this case social media contents, online news, comments). He will speak about the progress of how the lingustic algorithms and visualisation moduls together made a better and better composition gradually. He will also give an overview about the problems they had to face and feedbacks received from clients/users, and will present their unique indicators and charts (web opinion index, mention graph, attribute map).

Predictive analytics, and big data, in particular, offer the promise of better business decisions without relying on untested hypotheses. In this talk I argue that there is a role for economic theory in setting up business analytics projects and implementing subsequent business decisions. I first show some examples where purely data-driven decisions have gone awry. The reason is that agents we are collecting data about respond to changes in incentives, and will change their behavior after a change in business practices. I propose a symbiotic approach to theory and analytics.
Detailed contents:
1. How not to design class composition? External validity of predicted results.
2. How did this song get on my last.fm? A Lucas critique of consumer recommendations.
3. A checklist can save your life. When simple is better.

R is often criticized for not being able to handle large amount of records, and besides the performance issues, R cannot handle data in a memory-efficient way. This talk will give a quick overview on how I sorted out these problems while preparing for the Big Data Challenge, and how I could process 500+ Gb of CSV files to generate a short video on the volume of telecommunication and transportation in Milan. The resulting animation tells the story of human activity targeted at an audience without any prior statistical or similar knowledge.

There are a few simple rules of how to make an understandable chart, and most probably we all know what these are. In this presentation the speaker will sum these up and show some non-trivial caveats that should be avoided in order to make a chart easily understandable by everyone.

We all face challenges once users dream about new reporting Initiatives. How to let these dreams flow, while ensuring data will be there once we would like to populate these reports?
In her keynote presentation Janka Erdőssy showcases the most hyped trends of data visualization using her extensive hands-on experience with market leaders of different business sectors. She will give and discuss meaningful insights to the following questions:

• At what point should advance analytics be implemented in the pilots?
• Which factors determine the service we need for maximizing ROI of our BI reporting?
• Can we create pilots without data?

Today’s data sets are so large and complex that analytical methods of the past won’t work anymore.
Do you know what makes the real difference between market leaders and market followers? BI technology and analytical culture.
Real business examples illustrating the best practices from capturing data through analysis to visualization.