Structure and power dynamics in labour flow and company control networks in the UK

Data Science master project by Áron Pap ’20

Droplets of dew collect on a spider web
Photo by Nathan Dumlao on Unsplash

Editor’s note: This post is part of a series showcasing Barcelona GSE master projects. The project is a required component of all Master’s programs at the Barcelona GSE.

Abstract

In this thesis project I analyse labour flow networks, considering both undirected and directed configurations, and company control networks in the UK. I observe that these networks exhibit characteristics that are typical of empirical networks, such as heavy-tailed degree distribution, strong, naturally emerging communities with geo-industrial clustering and high assortativity. I also document that distinguishing between the type of investors of firms can help to better understand their degree centrality in the company control network and that large institutional entities having significant and exclusive control in a firm seem to be responsible for emerging hubs in this network. I also devise a simple network formation model to study the underlying causal processes in this company control network.

figure

Conclusion and future research

Intriguing empirical patterns and a new stylized fact are documented during the study of the company control network, since there is suggestive evidence that the types and number of investors are strongly associated with how “interconnected” a firm is in the company control network. Based on the empirical data it also seems that the largest institutional investors mainly seek opportunities where they can have significant control without sharing it with other dominant players. Thus the most “interconnected”/central firms in the company control network are the ones who can maintain this power balance in their owner structure. 

The devised network formation model helps to better understand the potential underlying mechanisms for the empirically observed stylized facts about the company control network. I carry out numerical simulations, sensitivity analysis and also calibrate parameters of the model using Bayesian optimization techniques to match the empirical results. However, these results could be “fine-tuned” at different stages further, in order to have a better empirical fit. First, the network formation model could be enhanced to represent more complex agent interactions and decisions. But also, the model calibration method could be extended to include more parameters and a larger valid search space for each of those parameters.

This project could also benefit from improvements to the utilised data. For example more granular data on the geographical regions could help to understand the different parts of London more and to have a more detailed view of economic hubs in the UK. Moreover, the current data source provides a static snapshot of the ownership and control structure of firms. Panel data on this front could enhance the analysis of the company control network, numerous experiments related to temporal dynamics could be carried out, for example link prediction or testing whether investors follow some kind of “preferential attachment” rules when acquiring significant control in firms.

Connect with the author

Áron Pap, Visiting Student at The Alan Turing Institute

About the Barcelona GSE Master’s Program in Data Science

Scalable Inference for Crossed Random Effects Models

Data Science master project by Maximilian Müller ’20

Editor’s note: This post is part of a series showcasing Barcelona GSE master projects. The project is a required component of all Master’s programs at the Barcelona GSE.

Abstract

Crossed random effects models are additive models that relate a response variable (e.g. a rating) to categorical predictors (e.g. customers and products). They can for example be used in the famous Netflix problem, where movie ratings of users should be predicted based on previous ratings. In order to apply statistical learning in this setup it is necessary to efficiently compute the Cholesky factor L of the models precision matrix. In this paper we show that for the case of 2 factors the crucial point to this end is not only the overall sparsity of L, but also the arrangement of non-zero entries with respect to each other. In particular, we express the number of flops required for the calculation of L by the number of 3-cycles in the corresponding graph. We then introduce specific designs of 2-factor crossed random effects models for which we can prove sparsity and density of the Cholesky factor, respectively. We confirm our results by numerical studies with the R-packages Spam and Matrix and find hints that approximations of the Cholesky factor could be an interesting approach for further decrease of the cost of computing L.

Key findings

  • The number of 3-cycles in the fill graph of the model are an appropriate measure of the computational complexity of the Cholesky decomposition.
  • For the introduced Markovian and Problematic Design we can prove sparsity and density of the Cholesky Factor, respectively.
  • For precision matrices created according to a random Erdös-Renyi-scheme the Spam algorithms could not find an ordering that would be significantly fill-reducing. This indicates that it might be hard or even impossible to find a general ordering rule that leads to sparse Cholesky factors.
  • For all observed cases, many of the non-zero entries in the Cholesky factor are either very small or exactly zero. Neglecting these small or zero values could spare computational cost without changing the Cholesky factor ‘too much’. Approximate Cholesky methods should therefore be included in further research.
Fill-in-ratio (a measure of relative density of the Cholesky factor) vs. matrix size for the random Erdös-Renyi scheme. For all permutation algorithms the fill-in-ratio grows linearly in I indicating that in general it might be hard to find a good, fill-reducing permutation.

Connect with the author

About the Barcelona GSE Master’s Program in Data Science

Machine Learning for the Sustainable Management of Main Water Supply Assets

Maryam Rahbaralam ’19 (Data Science)

big data

Maryam Rahbaralam ’19 (Data Science) presented “Machine Learning for the Sustainable Management of Main Water Supply Assets” with Jaume Cardús (Aigües de Barcelona) during the Pioneering Fields and Applications (Strong AI) session at the 2019 Big Data and AI Congress in Barcelona.

Abstract

The developed machine learning model gives the prediction of the probability of failure for each pipe section of the water supply network, allowing an early renewal of those in more detrimental conditions in terms of social, environmental and economic consequences.

Video

Maryam Rahbaralam ’19 is a Data Scientist at the Barcelona Supercomputing Center (BSC). She is an alum of the Barcelona GSE Master’s in Data Science.

LinkedIn | Twitter

Solving data science problems with Record Matching

Presentation by Data Science alum Jordan McIver ’15

presentation

Every organisation needs to be able to properly connect disparate datasets to take full advantage of their data assets. Alchemmy held an event to discuss approaches and technologies to connect datasets and watchouts to consider once they are connected.

Check out my talk here where we look at an approach that best enables data scientists by partnering them with the other staff who actually hold the context of the data:

Video summary

Most businesses have some or all of the following problems: not enough data science resources for the work required; a large community of data-adjacent staff who have most of the context but are not contributing what they know in the right way; data science problems lacking that same context; algorithms that cannot overcome a lack of data quality or availability of training data. Jordan walks through the use of interactive dashboards where users quality assess the data and this feeds back into the data science process which addresses these problems.

Jordan

Jordan McIver ’15 is Head of Data Consulting at Alchemmy in London. He is an alum of the Barcelona GSE Master’s in Data Science.

LinkedIn

Investigation of Sentiment Importance on Intraday Stock Returns

Data Science master project by Michele Costa, Alessandro De Sanctis, Laurits Marschall and S. Hamed Mirsadeghi ’18

Investigation of Sentiment Importance on Intraday Stock Returns

Editor’s note: This post is part of a series showcasing Barcelona GSE master projects by students in the Class of 2018. The project is a required component of every master program.


Authors:

Michele CostaAlessandro De SanctisLaurits Marschall and S. Hamed Mirsadeghi

Master’s Program:

Data Science

Paper Abstract:

The main goal of our Master Project is to predict intraday stock market movements using two different kinds of input features: financial indicators and sentiments from news and tweets. While the former are part of the common technical analysis of financial econometric models, the extracted sentiment of news articles and tweets from Twitter are also proven to correlate with stock markets movements. Our paper aims at contributing to the existing academic and professional knowledge in two main directions. First, we evaluate three different approaches to extract the sentiment from both social and mass media based on its forecasting power. Second, we deploy a battery of engineered features based on the sentiment, together with the financial indicators, in a machine learning model for a fine-grained minute-level forecasting exercise. In the end, two different classes of models are fitted to test the forecasting power of the combined input features. We estimated a classical ARIMA-model, and an XGBoost-model as machine learning algorithm. We collected data on the companies Apple, JPMorgan Chase, Exxon Mobil, and Boeing.

Figure: Exxon Mobil
The picture shows how sentiments towards Exxon Mobil moved over time. The two lines refers to two different methodologies: Loughran-McDonald is based on a financial dictionary while SentiStrength was trained on social media such as MySpace.


More about the Data Science Program at the Barcelona Graduate School of Economics

BGSE Data Talks: Professor Piotr Zwiernik

The Barcelona GSE Data Science student blog has a new post featuring an interview with Piotr Zwiernik (UPF and BGSE), Data Science researcher and professor in the BGSE Data Science Master’s Program.

The Barcelona GSE Data Science student blog has a new post featuring an interview with Piotr Zwiernik (UPF and BGSE), Data Science researcher and professor in the BGSE Data Science Master’s Program:

Hello and welcome to the second edition of the „Data Talks“ segment of the Data Science student blog. Today we have the honor to interview Piotr Zwiernik, who is assistant professor at Universitat Pompeu Fabra. Professor Zwiernik was recently awarded the Beatriu de Pinós grant from the Catalan Agency for Management of University and Research Grants. In the Data Science Master’s Program he teaches the maths brush-up and the convex optimization part of the first term class „Deterministic Models and Optimization“. Furthermore, he is one of the leading researchers in the field of Gaussian Graphical Models and algebraic statistics. We discuss his personal path, the fascination for algebraic statistic as well as the epistemological question of low-dimensional structures in nature…

Read the full interview on the Barcelona GSE Data Scientists blog

BGSE represented by “Just Peanuts” at Data Science Game finals in Paris

Class of 2017 Data Science graduates Roger Garriga, Javier Mas, Saurav Poudel, and Jonas Paul Westermann qualified for the final round of the Data Science Game in Paris this fall. Here is their account of the event.


Data Science Game is an annual competition organized by an association of volunteers from France. After competing in a tough online classificatory phase during the master we classified to the finals in Paris where we would be presented with a new problem to solve in a 2 days hackathon.

The hackathon was held in a palace property of Capgemini called Les Fontaines. It was an amazing building that made the experience even better.

The problem presented was to estimate the demand of 1.500 different products on 4 different countries using historic orders from 100.000 customers during the past 5 years by forecasting the three subsequent months. This was a well defined challenge that could be tackled with a large variety of solutions and for us specially the time constrain was one of the main challenges, since at the end we could be only 3 instead of 4.

We started by exploring the data and we realised that there were a lot of missing values due to a cross of databases done by the company who provided the data. So we spent some time by cleaning up the data and filling some of the missing values, to later on apply our models. After all the cleaning the key element to solve the challenge was later on to engineer good features that would represent well the data and then apply a simple model to predict the 3 months ahead.

The hackathon can be summed up in a day and a half coding, modeling and discussing without sleeping surrounded by 76 other participants from all across the world that were basically doing exactly the same, with short pauses to eat pizza, hamburgers and Indian food. So, a pretty good way to spend a weekend.

This slideshow requires JavaScript.

BGSE “Just Peanuts” qualifies for Data Science Game finals in Paris

A team of Barcelona GSE Data Science students from the Class of 2017 will compete in the final round of the Data Science Game in Paris at the end of September. 

data science game

A team of Barcelona GSE Data Science students from the Class of 2017 will compete in the final round of the Data Science Game in Paris at the end of September.

Among 400 international teams from 220 universities that participated in the first round, the BGSE team is among the 20 teams who have qualified for the final. The team is called “Just Peanuts” and its members are Roger Garriga, Javier Mas, Saurav Poudel, and Jonas Paul Westermann.

In the following interview, they talk about the Data Science Game and their expectations for the final.

What is the Data Science Game?

The Data Science Game is an annual Data Science competition for university students organized by ENSAE (Paris). Teams of up to four people can participate and represent their university. There is a free-for-all qualification round online and the top 20 teams are invited to the Finale in Paris.

Why did you decide to participate?

During the course we already took part in one data science challenge as part of the Computational Machine Learning course. That was quite fun and we have been generally wanting to take part in Kaggle-like challenges throughout the year. On top of that, we of course need to represent the Barcelona GSE and put the word out about our amazing Master’s.

Can you explain the task your team had to perform in the first round of the game?

The challenge for the online qualification round was related to predicting user’s music preferences. Data was provided by Deezer, a Music streaming service based in France. The training dataset consisted of 7+ million rows each pertaining to one user-song interaction describing weather the user listened to the song (for longer that 30 seconds) or not and whether the song was suggested to the user by the streaming service as well as further variables relating to the song/user.

How/by whom was the first round judged/scored?

The online round was hosted on Kaggle, a common website for these kinds of data science prediction challenges. Scoring was done according to the ROC AUC metric (reciever operator characteristic Area under the curve).

Was it difficult to combine participating in the game with your courses and assignments in the master program?

As we started really investing time into the challenge only quite late (about two weeks before the end) we spent a lot of time during the final days. The last 120 hours before submission were probably entirely spent on the challenge which definitely cut into our normal working schedules. Especially the last weekend before the deadline was very intense and spent mostly sitting shirtless at the table of a very overheated apartment living off frozen pizza and chips.

What specifically from the master’s helped you succeed in the game?

Part of the final model we used and what also made the first miles in terms of achieving a good score was a library recommended by one of the PhD students who also give lectures in our course. But also beyond that, we used all kinds of background knowledge and experience gained from the course. A constant scheme during the challenge were problems with difference in distribution and construction of the training and testing datasets. This gave inaccurately high cross-validation results and made it difficult to assess the quality of predictions.

Another issue was simply the size of the data that meant training and parameter tuning were extremely time consuming and we needed to expand our infrastructure beyond our own laptops. For both of those problems we’ve talked about possible solutions during the Master’s and applied combinations thereof.

What will you have to do for the final round? Can you tell us about your strategy or will that give too much information to the other teams?

The final round will be a two-day hackathon-like data science challenge on-site in Paris. No information has been shared with us on details of the challenge but we are thinking it might be something related to sound processing to continue the theme from part one.

How can we follow your progress in the competition?

We will surely be writing an update after the Paris trip and probably also give some social media updates during the event.

#ICYMI on the BGSE Data Science blog: Prediction as a Game

In this article we provide a general understanding of sequential prediction, with a particular attention to adversarial models.

Prediction as a Game

by Davide Viviano ’17

In this article we provide a general understanding of sequential prediction, with a particular attention to adversarial models. The aim is to provide theoretical foundations to the problem and discuss real life applications…

#ICYMI on the BGSE Data Science blog: RandNLA for LS (Part 2)

Randomized Numerical Linear Algebra for Least Squares – Part 2

by Robert Lange ’17

In today’s article we are going to introduce the Fast Johnson Lindenstrauss Transform (FJLT). This result is going to be the fundament of two very important concepts which speed up the computation of an ε-approximation to the LS objective function and the target vector…

See also Part 1 of this post