Tag: scraperwiki

Jan 20 2014

Book review: Mining the Social Web by Matthew A. Russell

mining_the_social_web_cover

This review was first published at ScraperWiki.

The twitter search and follower tools are amongst the most popular on the ScraperWiki platform so we are looking to provide more value in this area. To this end I’ve been reading “Mining the Social Web” by Matthew A. Russell.

In the first instance the book looks like a run through the APIs for various social media services (Twitter, Facebook, LinkedIn, Google+, GitHub etc) but after the first couple of chapters on Twitter and Facebook it becomes obvious that it is more subtle than that. Each chapter also includes material on a data mining technique; for Twitter it is simply counting things. The Facebook chapter introduces graph analysis, a theme extended in the chapter on GitHub. Google+ is used as a framework to introduce term frequency-inverse document frequency (TF-IDF), an information retrieval technique and a basic, but effective, way to process natural language. Web pages scraping is used as a means to introduce some more ideas about natural language processing and summarisation. Mining mailboxes uses a subset of the Enron mail corpus to introduces MongoDB as a document storage system. The final chapter is a twitter cookbook which includes lots of short recipes for simple twitter related activities but no further analysis. The coverage of each topic isn’t deep but it is practical – introducing the key libraries to do tasks. And it’s alive with suggests for further work, and references to help with that.

The examples in the book are provided as IPython Notebooks which are supplied, along with a Notebook server on a virtual machine, from a GitHub repository. IPython notebooks are interactive Python sessions run through a browser interface. Content is divided into cells which can either be code or simple descriptive text. A code cell can be executed and the output from the code appears in an output cell. These notebooks are a really nice way to present example code since the code has some context. The virtual machine approach is also a great innovation since configuring Python libraries and the IPython server itself, in a platform agnostic manner, is really difficult and this solution bypasses most of those problems. The system makes it incredibly easy to run the example code for yourself, almost too easy in fact, I found myself clicking blindly through some of the example code. Potentially the book could have been presented simply as an IPython notebook, this is likely not economically practical but it would be nice to collect the links to further reading there where they would be more usable. The GitHub repository also provides a great place for interaction with the author: I filed a couple of issues regarding setting the system up and he responded unerringly quickly – as he did for many other readers. Also I discovered incidentally, through being subscribed to the repository, that one of the people I follow on Twitter (and a guest blogger here) was also reading the book. An interesting example of the social web in action!

Mining the social web covers some material I had not come across in my earlier machine learning/ data mining reading. There are a couple of chapters containing material on graph theory using data from Facebook and GitHub data. In the way of benefitting from reading about the same material in different places, Russell highlights that cluster and de-duplication are of course facets of the same subject.

I read with interest the section on using a MongoDB database as a store for tweets and other data in the form of JSON objects. Currently I am bemused by MongoDB. The ScraperWiki platform uses it to store user profile information. I have occasional recourse to try to look things up there. I’ve struggled to see the benefit of MongoDB over a SQL database. Particularly having watched two of my colleagues spend a morning working out how to do a what would be a simple SQL join in MongoDB. Mining the social web has made me wonder about giving MongoDB another chance.

The penultimate chapter is a discussion of the semantic web, introducing both microformats as well as RDF technology, although the discussion is much less concrete than earlier chapters. Microformats are HTML elements which hold semantic information about a page using an agreed schema, to give an example: the geo microformat encodes geographic information. In the absence of such a microformat, geographic information such as latitude and longitude could be encoded in pretty much any way, making it necessary to either use custom scrapers on a page by page basis or complex heuristics to infer the presence of such information. RDF is one of the underpinning technologies for the semantic web: a shorthand for a worldwide web marked up such that machines can understand the meaning of webpages. This touches on the EU Newsreader project on which we are collaborators, and which seeks to generate this type of semantic mark up for news articles using natural language processing.

Overall, definitely worth reading. We’re interested in extending our tools for social media and with this book in hand I’m confident we can do it and be aware of more possibilities.

Jan 01 2014

Book review: Data Mining – Practical Machine Learning Tools and Techniques by Witten, Frank and Hall

datamining

This review was first published at ScraperWiki.

I’ve been doing more reading on machine learning, this time in the form of Data Mining: Practical Machine Learning Tools and Techniques by Ian H. Witten, Eibe Frank and Mark A. Hall. This comes by recommendation of my academic colleagues on the Newsreader project, who rely heavily on machine learning techniques to do natural language processing.

Data mining is about finding structure in data, the algorithms for doing this are found in the field of machine learning. The classic example is Iris flower dataset. This dataset contains measurements of parts of a flower for three different species of Iris, the challenge is to build a system which classifies a flower to its species by its measurements. More practical examples are in the diagnosis of machine faults, credit assessment, detection of oil slicks, customer support analysis, marketing and sales.

Previously I’ve reviewed Machine Learning in Action by Peter Harrington. Data Mining is a somewhat different book. The core contents are quite similar: background to machine learning, evaluating your results and a run through the core algorithms. Machine Learning in Action is a pretty quick run through the field touching on many subjects, with toy demonstrations built from scratch in Python. Data Mining, running to almost 600 pages, is a much more thorough reference. There is a place for both types of book, even on the same bookshelf.

Data Mining is written by three members of the University of Waikato’s Computer Science Department and is based around the Weka machine learning system developed there. Weka is a complete framework, written in Java, which implements the algorithms described in the book as well as some others. Weka can be accessed via the command-line or using a GUI. As well as the machine learning algorithms there are systems for preparing data, evaluating and visualising results. A collection of well-known demonstration data sets are included. I’ve no reason to doubt the quality of the implementations in Weka, the GUI interface is functional, occasionally puzzling and not particularly slick. The book stands alone from the Weka framework but the framework provides a good playground to try out the techniques discussed in the book. Weka seems to be entirely suitable for conducting serious analysis. This approach is in contrast to the approach of Harrington in Machine Learning in Action who provides toy implementations of algorithms in Python.

The first two parts of the book provide an overview of machine learning, followed by a more detailed look at how the key algorithms are implemented. The third section is dedicated to Weka, whilst the first two sections refer to it but do not rely on it. The third section is divided into a discussion of Weka, covering all its key features and then a tutorial. I found this a bit confusing since the first part has the air of a tutorial, but isn’t, and the tutorial part keeps referring back to the overview section for its screenshots.

With some knowledge already in machine learning, the things I learned from this book:

  • better methods, and subtleties in measuring the performance of machine learning algorithms;
  • the success of the one-rule algorithm, essentially a decision tree which gets the maximum benefit from a single rule. It turns such an approach is surprisingly effective and only bettered a little, if at all, by more sophisticated algorithms;
  • getting enough, clean data to do machine learning is often a problem;
  • where to learn more!

The first edition of this book was published in 1999; my review is of the third edition. The book does show some signs of age, Machine Learning in Action  was written as a response to a poll published at the International Conference on Data Mining 2006 on the 10 most important machine learning algorithms (see the paper here). Whilst Data Mining mentions this survey, it is as something of an afterthought and the authors seem bemused by the inclusion of the PageRank algorithm used by Google to rank web pages in search results. They mention the Moa framework for data stream mining although do not discuss it in any detail. Moa focuses on techniques for large datasets.

In summary: a well-written, well-structured and readable book on machine learning algorithms with demonstrations based on an extensive machine learning framework. Definitely one to read and come back to for reference.

Dec 31 2013

The BIG Lottery Data

uklogo

This post was originally published at ScraperWiki.

The UK’s BIG Lottery Fund recently released its grant data since 2004 as a set of lovely CSV files: You can get it yourself here or here. I found it a great opportunity to try out some new tricks with Tableau, and have a bit of a poke around another largish dataset from government. The data runs to a little under 120,000 lines.

The first question to ask is: where is all the money going?

The total awarded is £5,277,058,180 over nearly 10 years. It’s going to 81,386 different organisations. The sizes of grants vary enormously; the biggest, £214,340,846, going to the Big Local Trust, which is an umbrella organisation. Other big recipients include the Royal Society of Wildlife Trusts, who received £59,842,400 for the Local Food programme. The top 10 grants are listed below:

01/03/2012, Big Local Trust  £        214,340,846
15/08/2007, Royal Society of Wildlife Trusts  £          59,842,400
04/10/2007, The Federation of Groundwork Trusts  £          58,306,400
13/05/2008, Sustrans Limited  £          49,980,908
11/10/2012, Life Changes (Trustee) Limited  £          49,338,186
13/12/2011, Forces In Mind Trustee Limited  £          34,808,423
19/10/2007, Natural England  £          30,113,200
01/05/2007, Legacy Trust UK Limited  £          28,850,000
31/07/2007, Sustrans Limited  £          25,023,084
09/04/2008, Falkirk Council  £          25,000,000

Awards like this make determining the true geographic distribution of grants a bit tricky, since they are registered as being awarded to a particular local area – apparently the head office of the applicant – but they are used nationally. There is a regional breakdown of where the money is spent but this classification is to large areas i.e. “England” or “North West”. The Big Local Trust, Life Changes and Forces in Mind are all very recently established – less than a couple of years old. The Legacy Trust was established in 2007 to fund programmes to promote an Olympic legacy.

These are really big grants, but what does the overall distribution of awards look like?

This is shown in the chart below:

Award distribution

It’s a bit complicated because the spread of award sizes is from about £1000 to over £100,000,000 so what I’ve done is taken the logarithm of the award to create the bins. This means that the column marked “3” contains the sum of all awards from £1000 to £9999 and that marked “4” contains the sum of all awards from £10,000 to £99,999. The chart shows that most money is distributed in the column marked “5”, i.e. £100,000 to £999,999. The columns are coloured by the year in which money was awarded, so we can see that there were large grants awarded in 2007 as well as 2011 and 2013.

Everybody loves a word cloud, even though we know it’s not good in terms of data visualisation, a simple bar chart shows the relative frequency of words more clearly. The word cloud below shows the frequency of words appearing in the applicant name field of the data, lots of money going to Communities, Schools, Clubs and councils.

image

The data also include the founding date for the organisations to which money is awarded, most of them were founded since the beginning of the 20th century. There are quite a few schools and local councils in the list and, particularly for councils we can see the effect of legislation on the foundation of these organisations, there are big peaks in founding dates for councils in 1894 and in 1972-1974, coinciding with a couple of local government acts. There’s a dip in the foundation of bodies funded by the BIG lottery for both the First and Second World Wars, I guess people’s energies were directed elsewhere. The National Lottery started in the UK in late 1994.

Founding year

As a final piece of analysis I thought I’d look at sport; I’m not particularly interested in sport so I let natural language processing find sports for me in applicant names – they are often of the form “Somewhere Cricket/Rugby/Tennis/etc Club”. One way of picking out all the sports awards would be to come up with a list of sports names and compare against that list but I applied a little more cunning: the nltk library will tell you how closely related two words are using the WordNet lexicon which it contains. So I identified sports by measuring how closely related a target word was from the word “sport”. This got off to a shaky start  since I decided to use “cricket” as a test word; “cricket” is as closely related to “sport” as “hamster” – a puzzling result until I realised that the first definition of “cricket” in WordNet relates to the insect! This confusion dispensed with finding all the sports mentioned in the applicant names was an easy task. The list of sports I ended up with was unexceptional.

You can find participation levels in various sports here, I plotted them together with numbers of awards. Sports near the top left have relatively few awards given the number of participants, whilst those bottom right have more awards than would be expected from the number of participants.

 

Number of clubs vs number of participants

You can see interactive versions of these plots, plus a view more here on Tableau Public.

That’s what I found in the data – what would interest you?

Footnotes

I uploaded the CSV files to a MySQL database before loading into Tableau, I also did a bit of work in Python using the pandas library. In addition to the BIG lottery data I pulled in census data from the ONS and geographic boundary data from Tableau Mapping. You can see all this unfolding on the bitbucket repo I set up to store the analysis. Since Tableau workbook files are XML format they can be usefully stored in source control.

Nov 01 2013

Book review: Tableau 8 – the official guide by George Peck

tableau 8 guideThis review was first published at ScraperWiki.

A while back I reviewed Larry Keller’s book The Tableau 8.0 Training Manual, at the same time I ordered George Peck’s book Tableau 8: the official guide. It’s just arrived. The book comes with a DVD containing bonus videos featuring George Peck’s warm, friendly tones and example workbooks. I must admit to being mildly nonplussed at receiving optical media, my ultrabook lacking an appropriate drive, but I dug out the USB optical drive to load them up. Providing an online link would have allowed the inclusion of up to date material, perhaps covering the version 8.1 announcement.

Tableau is a data visualisation application, aimed at the business intelligence area and optimised to look at database shaped data. I’m using Tableau on a lot of the larger datasets we get at ScraperWiki for sense checking and analysis.

Colleagues have noted that analysis in Tableau looks like me randomly poking buttons in the interface. From Peck’s book I learn that the order in which I carry out random clicking is important since Tableau will make a decision on what you want to see based both on what you have clicked and also its current state.

To my mind the heavy reliance on the graphical interface is one of the drawbacks of Tableau, but clearly, to business intelligence users and journalists, it’s the program’s greatest benefit. It’s a drawback because capturing what you’ve done in a GUI is tricky. Some of the scripting/version control capability is retained since most Tableau files are in plain XML format with which a little fiddling is tacitly approved by Tableau – although you won’t find such info in The Official Guide. I’ve been experimenting with using git source control on workbook files, and it works.

If you’re interested in these more advanced techniques then the Tableau Knowledgebase is worth a look. See this article, for example, on making a custom colour palette. I also like the Information Lab blog, 5 things I wish I knew about Tableau when I started and UK Area Polygon Mapping in Tableau. The second post covers one of the bug-bears for non-US users of Tableau: the mapping functionality is quite US-centric.

Peck covers most of the functionality of Tableau, including data connections, making visualisations, a detailed look at mapping, dashboards and so forth. I was somewhat bemused to see the scatter plot described as “esoteric”. This highlights the background of those typically using Tableau: business people not physical scientists, and not necessarily business people who understand database query languages. Hence the heavy reliance on a graphical user interface.

I particularly liked the chapters on data connections which also described the various set, group and combine operations. Finally I understand the difference between data blending and data joining: joining is done at source between tables on the same database whilst blending is done on data from different sources by Tableau, after it has been loaded. The end result is not really different.

I now understand the point of table calculations – they’re for the times when you can’t work out your SQL query. Peck uses different language from Tableau in describing table calculations. He uses “direction” to refer to the order in which cells are processed and “scope” to refer to the groups over which cell calculations are performed. Tableau uses the terms “addressing” and “partitioning” for these two concepts, respectively.

Peck isn’t very explicit about the deep connections between SQL and Tableau but makes sufficient mention of the underlying processes to be useful.

It was nice to see a brief, clear description of the options for publishing Tableau workbooks. Public is handy and free if you want to publish to all. Tableau Online presents a useful halfway house for internal publication whilst Tableau Server gives full flexibility in scheduling updates to data and publishing to a range of audiences with different permission levels. This is something we’re interested in at ScraperWiki.

The book ends with an Appendix of functions available for field calculations.

In some ways Larry Keller and George Peck’s books complement each other, Larry’s book (which I reviewed here) contains the examples that George’s lacks and George’s some of the more in depth discussion missing from Larry’s book.

Overall: a nicely produced book with high production values, good but not encyclopedic coverage.

Oct 29 2013

Git!

logo@2x

This post was first published at ScraperWiki.

As software company, use of some sort of software source control system is inevitable, indeed our CEO wrote TortoiseCVS – a file system overlay for the early CVS source control system. For those uninitiated in the joys of software engineering: source control is a system for recording the history of file revisions allowing programmers to edit their code, safe in the knowledge that they can always revert to a previous good state of code if it all goes horribly wrong. We use Git for source control, hosted either on Github or on Bitbucket. The differing needs of our platform and data services teams fit the payment plans of the two different sites.

Git is a distributed source control system created by Linus Torvalds, to support the development of Linux. Git is an incredibly flexible system which allows you to do pretty much anything. But what should you do? What should be your strategy for collective code development? It’s easy to look up a particular command to do a particular thing, but less is written on how you should string your git commands together. Here we hope to address this lack.

We use the “No Switch Yard” methodology, this involves creating branches from the master branch on which to develop new features and regularly rebasing against the master branch so that when the time comes the feature branch can be merged into the master branch via a pull request with little fuss. We should not be producing a byzantine system by branching feature branches from other feature branches. The aim of “No Switch Yard” is to make the history as simple as possible and make merging branches back onto master as easy as possible.

How do I start?

Assuming that you already have some code in a repository, create a local clone of that repository:

git clone [email protected]:scraperwiki/myproject.git

Create a branch:

git checkout -b my-new-stuff

Start coding…adding files and committing changes as you go:

git add -u
git commit -m "everything is great"

The -u switch to git add simply checks in all the tracked, uncommitted files. Depending on your levels of paranoia you can push your branch back to the remote repository:

git push

How do I understand what’s going on?

For me the key revelation for workflow was to be able to find out my current state and feel pleasure when it was good! To do this, fetch any changes that may have been made on your repository:

git fetch

and then run:

git log --oneline --graph --decorate --all

To see an ASCII art history diagram for your repository. What you are looking for here is a relatively simple branching structure without too many parallel tracks and with the tips of each branch lined up between your local and the remote copy.
You can make an alias to simplify this inspection:

git config --global alias.lg 'log --oneline --graph --decorate'

Then you can just do:

git lg --all

I know someone else has pushed to the master branch from which I branched – what should I do?

If stuff is going on on your master branch, perhaps because your changes are taking a while to complete, you should rebase. You should also do this just before submitting a pull request to merge your work with the master branch.

git rebase -i

Allows you to rebase interactively, this means you can combine multiple commits into a single larger commit. You might want to do this if you made lots of little commits whilst achieving a single goal. Rebasing brings you up to date with another branch, without actually merging your changes into that branch.

I’m done, how do I give my colleagues the opportunity to work on my great new features?

You need to rebase against the remote branch onto which you wish to merge your code and then submit a pull request for your changes. You can submit a pull request from the web interface at Github or Bitbucket. Or you can use a command line tool such as hub.  The idea of using a pull request is that it makes your changes visible to your colleagues, and keeps a clear record of those changes. If you’ve been rebasing regularly you should be able to merge your code automatically.

An important principle here is “ownership”, in social terms you own your local branch on which you are developing a feature, so you can do what you like with it. The master branch from which you started work is in collective ownership so you should only merge changes onto it with the permission of your colleagues and ideally you want others to look at your changes and approve the pull themselves.

I started doing some fiddling around with my code and now I realise it’s serious and I want to put it on a branch, what did I do?

You need to stash your code, using:

git stash

Then create a branch, as described above, and then retrieve the contents of the stash:

git stash pop

That’s how we use git – what do you do?