49 Responses to “DM SIG – ACM Silicon Valley Data Mining Camp on November 1, 2009”

1. Martin Stein says:

   October 9, 2009 at 6:10 pm

   Please be advised that there is also

   NASA’s Conference on Intelligent Data Understanding (Oct 14-16, 2009)
   https://dashlink.arc.nasa.gov/group/ciduaisrp/
2. Scott Feamster says:

   October 12, 2009 at 9:07 pm

   Our proposed topic is: E-commerce Data Mining.
   How much time is available to us?

   Thank you very much for organizing the Data Mining Camp.
3. GregMakowski says:

   October 13, 2009 at 11:32 am

   Scott,
   E-commerce Data Mining sounds fine for a subject. There are 4 time slots, with about 50 minutes per time slot and 10 minutes for break and to switch to the next time slot. Each time slot starts on the hour.

   If you are interested in proposing more than one time slot, then I would suggest you propose variations on a theme. For example,
   * “E-commerce Data Mining: analysis for banner ad selection”
   * “E-commerce Data Mining: clustering customers to understand the best communication messages”

   Propose each topic, and if the threshold of people is interested in each session, then you can hold multiple sessions in the unconference format.

   Thanks,
   Greg_Makowski at yahoo.com
4. GregMakowski says:

   October 17, 2009 at 5:35 pm

   TOPIC: Bring your project challenges to brainstorm on current projects with experts (discussion)

   OBJECTIVE: Provide the project owner with new insights and approaches to their problem.

   DETAIL:
   * Do you want to get some help assessing if a business problem would be a good fit for analysis?
   * Do you want to discuss how to structure a problem, the business value metrics, how to explore or prepare the data?
   * Are you working on an existing project, the improving results have leveled off, and you want to brainstorm alternative approaches?

   Contact us in advance, or blog, if you have questions on how to best articulate your challenges. Providing detail in advance to read can help others think and possibly prepare.
5. GregMakowski says:

   October 17, 2009 at 5:48 pm

   TOPIC: Bring your business problems and data for a “data mining coopertition hackathon” (data analysis)

   OBJECTIVE:
   Advance the problem owner’s insight into their problem, provide analysts with an interesting and useful challenge.

   DETAIL:
   This session would need two groups of prepared people to be successful:

   1) PROBLEM OWNER(S): People with business problems, data and some articulation of the objective or success. Bring the data and specification on multiple memory sticks – or blog in advance and leave any web links to the data and/or specification. I would suggest comma or delimited data with field names in the first row if it is a structured data problem. See also http://www.kdnuggets.com/datasets/index.html for ideas and formats.

   2) ANALYSTS: …with portables and analytic software willing to dive into a problem on a brief basis.

   One way this session could proceed:
   a) intro to the overall problem(s) and context
   b) if a choice exists among problems, pick one or divide efforts
   b) problem owner goes intro to details
   c) analysts self-organe on how to cooperatively or competitively partition working on the problem (Exploratory Data Analysis, preprocessing, running it through various software, …)
   d) iteratively work, regroup and share insights

   If it is of interest to both groups (analytic and problem owner), this could become a track, spanning one or all time slots, allowing a more significant “cloudsourcing” of investigation effort.
6. GregMakowski says:

   October 17, 2009 at 5:57 pm

   TOPIC: Netflix $1,000,000 data mining competition, presentation of collaborative filtering papers by Yehuda Koren from Greg Makowski

   OBJECTIVE: Review some of the extensive literature on collaborative filtering, discuss as a group, relate to the next Netflix data mining competition.

   DETAIL:
   We can have a general discussion on the subjects covered by the following reading:

   * High Level Reading
   Matrix Factorization Techniques for Recommender Systems Yehuda Koren; Robert Bell; Chris Volinsky, IEEE Computer, 2009, 8

   Programming Collective Intelligence by Toby Segaran. The 2nd chapter gives a good introduction to collaborative filtering with Python examples.

   * Detailed Reading
   See papers on research.yahoo.com/Yehuda_Koren

   Collaborative Filtering with Temporal Dynamics, Yehuda Koren, KDD 2009, ACM, 2009

   Collaborative Filtering for Implicit Feedback Datasets, Yifan Hu; Yehuda Koren; Chris Volinsky, IEEE International Conference on Data Mining (ICDM 2008), IEEE, 2008

   Factorization Meets the Neighborhood: a Multifaceted Collaborative Filtering Model, Yehuda Koren, ACM Int. Conference on Knowledge Discovery and Data Mining (KDD’08), 2008
7. Irene Gabashvili says:

   October 18, 2009 at 3:38 pm

   Proposed topic: Biomedical Data Mining: Dimensionality, Noise, Applications
8. liana Ye says:

   October 20, 2009 at 4:44 am

   I’d like to propose a topic on semantic web:
   “Build structured tags from Wordnet”
   the main idea is discussed on
   http://lianaye.mysite.com
9. Mark Richard-Fogg says:

   October 20, 2009 at 7:57 pm

   Yes, I’ll attend.
10. Ben Gimpert says:

    October 22, 2009 at 12:33 pm

    TOPIC: Finance & Trading meets Data Mining & Machine Learning

    STRUCTURE:

    - How do we identify bull, bear and sideways markets, without the benefits of hindsight, and without resorting to chartist rhetoric? Consider market and macro- factors that define attractors in a continuous state space.

    - Discuss typical pitfalls when technologists present data mining trading strategies to fund managers.

    - How are machine learning models trained on data mined trading and market information, from a practitioner’s perspective?

    READING:

    Singular Spectrum Analysis, Elsner & Tsonis

    Inefficient Markets, Schleifer
11. Mike Bowles says:

    October 22, 2009 at 2:32 pm

    TOPIC: Financial Track – Dealing with the changing character of financial markets.

    Financial markets change character in ways that are difficult to characterize. Sometimes they move quickly, sometimes slowly, sometimes up, sometimes down, sometimes sideways etc. Sometimes markets change character in response to identifiable events, sometimes for no apparent reason.

    Can data mining extract useable information in this setting? How?
12. Mike Bowles says:

    October 22, 2009 at 3:20 pm

    TOPIC: Financial Track – Approaches to posing financial data mining problems.

    The character of financial markets makes it difficult to formulate well-posed problems. For example, problem statements like “give buy and sell signals” or “is action A better or is B” seem well-posed, but it turns out to be complicated to even generate the truth sets for training.

    What are the alternatives?
13. Paul O'Rorke says:

    October 22, 2009 at 6:49 pm

    TOPIC: Open Source Data Mining Tools

    OBJECTIVES: Identify (list) open source tools for data mining such as Weka and Mahout. Describe and evaluate them: what do they contain? What do they do and which applications are they best suited for?
14. Paul O'Rorke says:

    October 22, 2009 at 6:58 pm

    TOPIC: Open Data Sources

    OBJECTIVES: List sources of data that are openly available and describe their uses.

    DETAILS: In the old days, it was difficult to get access to large collections of data for various reasons but mostly because they were proprietary. Since the advent of the internet more and more data is becoming available online. Toby Segeran’s book “Collective Intelligence” has a great set of examples. He shows a set of simple applications and he shows how to get data from the web to develop and test the applications. The objective here would be to build on this list and create a larger list of data sources and talk about their uses.
15. Mark Richard-Fogg says:

    October 23, 2009 at 8:29 am

    It would be remiss not to include the viewpoint of Nick Taleb when discussing data mining, non?

    Perhaps (some other credible speaker might) mention of these two articles to which Taleb refers might be of use:

    Halbert White 2000, A Reality Check for Data Snooping, Econometrica, Vol. 68 (Sep.), No. 5., pp. 1097-1126.

    R. Sullivan, A. Timmermann, and H. White: “Data Snooping, Technical Trading Rule Performance, and the Bootstrap,” Journal of Finance, 54, 1647-1692 (1999).
16. Mike Bowles says:

    October 24, 2009 at 3:13 pm

    Financial Track – Financial Data Variability

    I put together a short white paper discussing the character of variability in financial data. I think it’s an easy read with lots of pictures. If you find any mistakes or lack of clarity, let me know. Here’s the link:

    http://docs.google.com/Doc?docid=0ARinKxOPitdhZGhycWJ6OWdfMWZzdGg3OWZn&hl=en
17. liana Ye says:

    October 24, 2009 at 5:37 pm

    Hello, Mike,

    I have tried your link and get a nice intro page on Google Docs, but not your article. Do you need to set up invitation email addresses for others to read?

    Topic proposal:
    Using game strategy, Bubble Shooter (Free download from DeadWhale.com ) for tag voting or organizing data mining software standards.
18. Mike Bowles says:

    October 24, 2009 at 11:37 pm

    Financial Track – Republished Data Variability

    I have the permissions right this time, I believe. Here’s the new link:
    http://docs.google.com/View?id=dhrqbz9g_8d4fjqdcc
19. Steve Umfleet says:

    October 26, 2009 at 1:56 pm

    The Financial Track seems to have attracted a lot of interest. There a tons of questions to address, and this doc (I hope the permissions are correct) poses many of them, and provides some definitions and motivation for trying to answer them.
    http://docs.google.com/present/edit?id=0AS789QYRolhSZGZoand2eF8xMGdoOXBya2Rj&hl=en
20. Ken Krugler says:

    October 26, 2009 at 7:33 pm

    TOPIC: Elastic web mining using open source (Hadoop/Cascading/Bixo) in Amazon’s EC2 cloud.

    OBJECTIVES: Explain how to leverage on-demand/cloud computing for fast, cost-effective web mining tasks.

    Cover common issues around web crawling, scaling, and reliable workflow.

    DETAILS: I’ll be describing the problem space in general, and one viable solution (Hadoop/Cascading/Bixo in EC2) in detail, using an example of search engine optimization (SEO) as the concrete use case.
21. BootstrappersBreakfast » Hacker Dojo: Something is Happening in Mountain View says:

    October 27, 2009 at 3:47 am

    [...] days for the dojo, but a lot is happening there. While we were touring the folks organizing the 2009 Silicon Valley Data Mining Camp that will be held at the dojo were also there making preparations. It’s an interesting [...]
22. Nicolas Cebron says:

    October 27, 2009 at 4:10 pm

    I would also be interested in
    TOPIC: Open Source Data Mining Tools

    I could give a presentation of the new version of KNIME (http://www.knime.org)

    —————
    Nicolas,
    Yes, that would of interest as a subject. We can discuss Sunday if it makes sense to join a session or be a separate session. If you are able to bring a computer projector, that could be helpful.
    Thanks, Greg Makowski
23. Presenting at 2009 Silicon Valley Data Mining Camp « Elastic Web Mining | Bixolabs says:

    October 30, 2009 at 10:42 am

    [...] I’ll be presenting at some point during the day – since it’s an unconference, you don’t really know who’s going to be talking about what/when. My topic is “Elastic web mining using open source (Hadoop/Cascading/Bixo) in Amazon’s EC2 cloud“. [...]
24. Patricia Hoffman says:

    October 31, 2009 at 4:27 am

    For all of you that have been suggesting topics, I have created sheets with your name and topic title. These can be used at the beginning of the conference to help organize which topics will be held in which rooms at what time slot.

    Others that haven’t blogged yet are still welcome to suggest topics. Paper will be available for you to create sheets when you arrive.

    If you are facilitating a topic, it would be great if you could find a person to help be a scribe and another person to help with the time keeping of your session.

    Let the excitement begin … Patricia Hoffman
25. Ken Krugler says:

    November 1, 2009 at 7:30 pm

    =============================
    Open Source Data Mining Tools
    =============================

    Paul O’Rorke talked about Weka, a collection of machine learning algorithms for data mining tasks. Concerns about whether it’s still viable. One person said that pieces of it are still viable for clustering, feature selection.

    An attendee mentioned MOA. MOA is a framework for data stream mining. Includes tools for evaluation and a collection of machine learning algorithms. Related to the WEKA project, also written in Java, while scaling to more demanding problems.

    David talked about R. Possible to quickly get results by using building blocks from other users. Often data is prepared before processing by R. On the back end is presentation tools. Sweave is a report generation backup that works well with R. Lots of research going on for out-of-memory modeling, to handle larger data sets. Also lots of work in parallel processing. BigMemory is a package for large models. Paul mentioned that R has a steep learning curve. David agreed that R is quirky, especially in terms of memory usage.

    Attendee asked about comparing Matlab & R, with respect to viability in a production environment. He’d run into memory problems with Matlab. David said that it was similar, and recommended doing scoring outside of R. He estimates 3-6x more memory is required for R vs. C++.

    Attendee said many people use R for prototyping and generating models, but production uses something else. Examples would be Numpy and SciPy.

    Paul mentioned that R provides a very compact representation of data mining tasks. (Ken – so it’s the APL of data mining?)

    Nicolas Cebron talked about KNIME (pronounced “naim”), a modular data exploration platform. Started in 2004. knime.org has full details. He demonstrated the KNIME application, which has a nice GUI for working with data sets. The model can be output as PMML.

    Attendee asked about long-term viability of KNIME. Nicolas said that it’s been around for 4 years, has a vibrant community, and there are commercial companies creating modules.

    Ted Dunning talked about Mahout, an Apache open source project with the goal of scalable machine learning/data mining. Java is main language, Hadoop & Lucene are foundation technologies. Currently has good algorithms for clustering, kmeans. Reasonably good classifiers. Supervised learning algorithm. Also recommendation framework called TASTE. Very young project. Has support for sparse matrix math – might pool efforts with Apache commons math project. Mahout is mature enough for some types of machine learning problems.

    Chris Wensel from Concurrent talked about the Hadoop distributed file system, and how it differs from Sun’s distributed file system – HDFS is very specialized, optimized for streaming reads. Can’t do random updates to files. Scales to 1000s of servers. Very fault tolerant.

    Ken Krugler (your faithful scribe) talked about the HECB (Hadoop, EC2, Cascading, Bixo) stack for web mining. Focus is on the collection and initial processing/reduction of the data, not hard core machine learning & data mining.
26. ACM Data Mining Camp Sillicon Valley, report | Zemanta Ltd. says:

    November 1, 2009 at 9:19 pm

    [...] here’s the report from today’s ACM Data Mining Camp Silicon Valley. This is not exactly live blogging, but it is neither the deep thinking, so do further research and [...]
27. Andraz Tori says:

    November 1, 2009 at 9:23 pm

    Here’s my report from the camp [meant as internal company report initially]:

    http://www.zemanta.com/fruitblog/acm-data-mining-camp-sillicon-valley-report/

    It was fun, great organizers!

    bye
    Andraz Tori
28. Vipul Sharma says:

    November 2, 2009 at 12:48 pm

    I organized a session on Real Time Modeling challenges. There was a total of 12 attendees. Discussion was focused towards machine learning problems that require real time modeling. Discussion was divided into following parts:

    1) Data-
    i) Collection: which was agreed is less of a problem given the vast amount of real time data available from various sources. The

    ii) Bigger challenge is to clean data from noise in near real time.

    2) Mining:
    i) Automated feature generation: need to automatically generated features in near real time is challenging. Unsupervised techniques or clustering along with techniques like SVD are used. Attendee suggested using SVD before clustering to reduce clustering cost.

    ii) real time model training was also discussed. Difference between active training and real time training were discussed. Model validation and need to check the error rates on unseen data and checking for overfitting was also discussed as an important step.

    DJ Patil from linkedin questioned at whether retraining the models will be more benefitial than extracting new features. It was mentioned that in some cases classification errors are due to lack of new features than the decay of the model. Specific problems of spam and attendee from Google discussed click fraud problem in this regard.

    DJ Patil also discussed that buying sla time by showing users that system is trying to find the the best solution helps.

    Project http://hunch.net/~vw/ was mentioned.
    Other key discussion topic were K means clustering and how to find “K”. Canopy clustering, SVD/PCA. Attendee mentioned that for their problem ensemble techniques gave the best results.

    Attendees that I can remember are:

    Vipul Sharma, Proofpoint (sharmavipul AT gmail DOT com)
    Uri Roduy (varancon AT gmail DOT com)
    DJ Patil, linkedin
    Prayrana Khadye (pkhadye AT gmail DOT com)
    Pandu Rudraraju (pandur AT gmail DOT com)
    Roy Kamimura CODEXIS (roy DOT Kamimura AT codexis DOT com)
    Patrick R Nicolas, Semantic Web (patrick AT pnexpert DOT com)
    Ben Gimpert (ben AT somethingmodern DOT com)

    I thought event was a big hit. Lots of people from academia and industry showed up. Companies like twitter, linkedin, google, IM shopping etc were present. Lots of discussion on R. Large scale and near real time machine learning is obviously becoming a louder topic of discussion and hence hadoop, mahout, Bixo etc were mentioned a lot.

    Kudos to organizers!

    –
    Vipul Sharma
29. Ken Krugler says:

    November 2, 2009 at 4:45 pm

    And here’s my report on the open/public dataset discussion:

    http://bixolabs.com/datasets/public-datasets/

    I agree with Andraz – a very useful event, thanks!

    – Ken
30. Ken Krugler says:

    November 2, 2009 at 5:06 pm

    Also here’s a better version of my open source data mining tools report (posted above), with some links and cleaned up refs.

    http://bixolabs.com/oss/open-source-data-mining-tools/

    – Ken
31. Ken Krugler says:

    November 2, 2009 at 7:37 pm

    I also posted the Powerpoint of my talk on elastic web mining:

    http://www.slideshare.net/kkrugler/elastic-web-mining

    Same, but in PDF so you get the notes:

    http://www.slideshare.net/kkrugler/elastic-web-mining-2407818

    – Ken
32. Irene Gabashvili says:

    November 2, 2009 at 10:28 pm

    Report on biomedical/health-care data mining & Discussion transcript:

    http://aurametrix.blogspot.com/2009/11/biomedical-data-mining-dimensionality.html
33. Paul O'Rorke says:

    November 3, 2009 at 12:28 am

    I’ve posted notes on Ken Krugler’s session on elastic web mining using the HECB open source stack (Bixo, Cascading, Hadoop & EC2) at http://ororke.com/paul/blog/?p=261

    I don’t remember any controversial exchanges during the Q&A but if anyone remembers something important that I missed please feel free to post it there. Thx!
34. Patricia Hoffman says:

    November 3, 2009 at 9:50 am

    Press and Buzz

    Thank you all for coming. I appreciate everyone for helping out. It was a great conference. I hope that everyone had a great time.

    There is a blogs about the camp:

    General impression: http://www.zemanta.com/fruitblog/acm-data-mining-camp-silicon-valley-report/

    Focus on R: http://bit.ly/3sM5kQ
    BioMedical: http://aurametrix.blogspot.com/2009/11/biomedical-data-mining-dimensionality.html

    There were many “tweets” about the camp
    People even put up a twitpics
    http://twitpic.com/nx57f
    http://tweetphoto.com/xbs1abje
    http://img682.yfrog.com/i/b07.jpg/

    Here are some slides about the event! http://www.slideshare.net/clibou/datacamp
35. Steve Umfleet says:

    November 4, 2009 at 9:22 pm

    Summary of presentation by Monica Anderson

    Model-Free Methods
    2:30

    The presentation was based on http://artificial-intuition.com/

    Model Free methods

    Chaos theory, complexity theory, complex systems all cover some of the same ideas.

    Chaotic:
    Deep complexity
    Nonlinear responses
    State (memory)

    Irreducible systems:
    Open systems
    High dimensionality
    Time variance: interact with environment, internally modified

    Ambiguous input:
    Incomplete nput
    Incorrect input
    Multiple opinions
    Strange references

    Emergence:
    Quality of life, testing, maintenance

    Bizarre domains: have all the characteristics noted above re: chaotic, irreducible, ambiguous input, emergence

    World models, partial worlds models, weather, climate
    Life is bizarre: organisms genomics, human interactome

    Reductionist models don’t work in bizarre systems.

    People: societies, consumers,

    Brains, minds, intelligence,

    Language is bizarre:has all of the above characteristics

    AI Complete: once you solve any one of these problems you have in theory solved all of them.

    Chess: not bizarre: it’s a well-contained, well-defined system.

    Model-free methods: Not everything in science is done by models or methods.

    Reductionist science and everyday life are very distinct. Life sciences are part reductionist, part non-reductionist.

    Models are derived form low level observation‡ scientific intuition –> scientific model.

    Leibniz:

    Engineers use models: low level observation –> engineer’s intuition –> scientific model (preconditions) ‡ scientific predictions.

    Model free methods: subscientific intuition
    Low level observation –> experience & intuition –> intuitive prediction. In a bizarre domain you cannot create or generate a model.

    Easy to bizarre problems.
    Newtonian mechanics
    GPS navigation
    …
    language semantics

    Full scientific models
    Partial models (perilous)
    Simulations
    Statistics (depends on large numbers, distribution)
    Non-parametric models to figure out or determine the distribution
    Pseudo models
    Model free models

    The harder your problem the weaker your model and the more data you need.

    Required understanding: full domain understanding, decent data, know nothing.

    Evolution is a model free method.

    Model free method zoo:
    Language, evolution, adaptation, abstraction, learning, recognition, discovery, trial and error, repeat success, table lookup, consultation (very powerful).

    Memory helps for progression. Sensory input.

    Movie recommendations:
    Do not attempt to model the movies
    Don’t attempt to model the customers

    Big data: use weaker models: non-parametric models, pseudomodels, model free methods.

    Holism is opposite of reductionist methods.
    Simplify problems by reduction, reduce a whole to its parts, reduce context dependencies, reduce to next lower discipline, reduce your data, reduce wasted effort.

    A revolutionary capability:
    Reductionists are masters of methods
    Holists have superior ontology
    Holists: work on these important problems.

    Big data makes holistic methods possible
    Model free methods are holistic.

    If the problem is too hard, get MORE data.
    Requires large computers.
    Weaker models: let data speak for itself
    Used in genomics and other life sciences

    Reductionism: optimality, completeness, repeatability, timeliness, parsimony, economy, transparency, scrutability. Correctness, infallibility, portability, incrementality, idealism

    Holism: discovery, learning, self-organizing systems, saliency, abstraction. Robustness, flexibility, diversity, synergy, pragmatism.

    R: deduction, deduced novelty

    H: induction, abduction, true novelty.

    Summary:
    In many important but difficult problem domains, building reductionist models is impossible. We call these Bizarre Domains.

    In general the harder your domain the weaker your models, and the more data you need.

    Big data makes model free methods possible.
36. Steve Umfleet says:

    November 4, 2009 at 9:26 pm

    Notes (thanks to Greg) on Automated Trading Systems: What are the questions?

    3:30 What are the questions in financial trading? Steve Umfleet

    Slides at:
    http://docs.google.com/present/view?id=0AS789QYRolhSZGZoand2eF8xMGdoOXBya2Rj&hl=en

    Fundamental analysis (price should be)
    Technical analysis (price will be)

    Effect of news
    What does Thorpe do:
    Liquidate position; don’t trade anything that has a news event
    Trade off volatility: don’t necessarily care whether the news if favorable/unfavorable
    Internet has sped up recognition of bullish news on equities
    Less-well followed companies may have longer response periods

    A Random Walk Down Wall Street
    Sharpe: transaction costs make market effectively random
    Schiller: Irrational Exuberance
    What about seasonal effects? Momentum effects?

    Knowledge about how to exploit complexity in market diffuses
    Pairs trading worked for a while
    Index trading worked for a while

    Database scans: pattern matches. Brittle over time.
    Trading system: when to enter or exit stock
    Stock recommender system: identify a market situation and stocks to go long/short in. Diagnosing market state.
    Portfolio management system: cash, stocks, bonds.
    The moon

    Portfolio allocation theory: academic studies that say you do better by focusing on allocation rather than picking specific stocks or temporal trading
    Another academic study: Black-Scholes assumptions say that technical trading doesn’t work

    Sharpe assumes that there is a correct price for a stock. Question: is there a single correct price (chaotic attractor).

    Regime switching, timing, shifts in market timing: map to concepts in chaos, non-linear dynamics

    Individual investors need to work on smaller stocks
    Crossy

    Paradigms
    Neural networks
    SVM
    Decision trees
    SVD (see empirical orthogonal functions)
    Genetic algorithms
    Statistical filtering
    Fuzzy logic
    Production systems
    Will need to use multiple paradigms

    For what are you looking? Need some hypothesis about what creates mis-pricing?
    Tremendously high dimensionality and volume
    Awash in data

    Machine skills complement human skills
    Psychology of human sense systems
    Mirages exist in all sorts of human systems for perception
    Emotionality

    Tradestation (use forum, not doc), Neuroshell, SVMLight, libSVM, NeuralWare
    TreeNet
    NinjaTrader

    Opinion: trading system should be completely automatic, no room for emotion
37. Patricia Hoffman says:

    November 6, 2009 at 8:10 am

    More feed back from the camp can be found here: http://www.ai-meetup.org/calendar/11592442/
38. Patricia Hoffman says:

    November 8, 2009 at 11:01 am

    Web Mining in the Cloud

    Ken Krugler’s interesting presentation on elastic web data mining is discussed on Paul O’Roke’s blog (http://ororke.com/paul/blog/?p=261 ). Ken is the founder of Bixo Labs, Inc.

    The Bixo site (http://bixolabs.com/2009/11/04/paul-ororke-summary-of-elastic-web-mining-talk/ ) shows how by using Bixo and Cascading much of the work is done for you.
39. Patricia Hoffman says:

    November 8, 2009 at 11:03 am

    Ken Krugler’s talk can be found here

    http://bixolabs.com/2009/11/02/elastic-web-mining-talk/
40. Patricia Hoffman says:

    November 8, 2009 at 11:29 am

    Elastic Web Mining View more documents from Ken Krugler.
41. Patricia Hoffman says:

    November 8, 2009 at 1:55 pm

    Shamod Lacoul gave a Semantic Web Talk which can be seen here http://www.cliveboulton.com/post/231943312/semantic-web-overview-via-shamod-dmcamp
42. Patricia Hoffman says:

    November 8, 2009 at 1:57 pm

    Semantic Web demos can be seen here

    http://www.hewettresearch.com/svcc2009/

    A great text for beginners looking into the Semantic Web …

    “Explorer’s Guide to the Semantic Web” by Thomas B. Passin
43. Patricia Hoffman says:

    November 13, 2009 at 1:42 pm

    Sign up to be notified of the next Data Mining Camp

    http://tinyurl.com/acmdmsig
44. Greg Makowski says:

    December 10, 2009 at 6:31 pm

    Hello,
    I am starting to organize another Data Mining Camp, perhaps Sat 3/6 or 3/20.

    While Hacker’s Dojo is a great facility, with 225 people participating last time, and expectations of more people next time, I am investigating options for other locations. I am seeking contacts at large companies or universities, to discuss holding the next free Data Mining Camp.

    I am also seeking constructive suggestions for improving the next DM Camp – please send to my email (listed at the top of this posting).

    Thanks, Greg
45. Patricia Hoffman says:

    February 4, 2010 at 11:07 am

    To view videos of the conference look at the following web sites:

    video.syntience.com and vimeo.com

    Thanks to Syntience and Monica Anderson!
46. Patricia Hoffman says:

    February 6, 2010 at 5:03 pm

    Comments posted praising the November Conference!

    Patrick Nicolas
    “ I do not think that you can gather so many experts in this field on a Sunday afternoon anywhere in the world except Silicon Valley. The organization of the camp was just awesome. I was able to share both technical issues and potential solutions with some very smart engineers. Greg and Tricia did a remarkable job. ”

    Bill Tang
    “ This is a great camp. The organizer did a fantastic job. The panel and the people leading the sessions are some of the best people in this field. I wish I could have attended all the parallel sessions. ”

    Bob Blum
    “ Fantastic! Great lectures, bright participants, fascinating subject. ”
47. Greg Makowski says:

    March 14, 2010 at 12:35 am

    Here is the SESSION MATRIX from the 11/1/2009 Data Mining Camp.

    Please send any updates or corrections, and I will revise.
48. David Smith says:

    March 31, 2010 at 2:58 pm

    Thoughts on the SF Bay Data Mining Camp
    I had a great time at the Data Mining Camp hosted by the ACM yesterday. The event was full of energy — despite original projections of around 100 attendees, in fact more than 200 people from around the region turned up to meet and discuss various aspects of data mining. This was an “unconference” – except for a 30-minute panel discussion (great discussion from folks at Google and LinkedIn, amongst others) there was no pre-set agenda. Instead, people proposed topics for discussion, participants expressed interest with a show of hands, and the talks were allocated to timeslots and rooms accordingly.

    I wish I’d taken a photo of the papers stuck to the wall making up the final schedule — I’d guess in the end there were well over 20 talks on various topics related to data mining. Large-scale data processing with Hadoop and its machine-learning cousin Mahout was a hot topic. So was financial data mining (which was surprising to me, for a West Coast event). There were also talks on the semantic web, natural language processing, and many other topics I can’t remember now.

    I proposed a topic on Data Mining and Machine Learning with R, which with a show of hands about 50 people were interested in. Then someone proposed a topic on Basic R, and more than 80 people signed up for that one. So there was a lot of interest in R at the conference — lots of people had heard about R but hadn’t yet used it. I participated in both those sessions. For the Basics session, I have an introduction to R resources for beginners and the R syntax. For the Machine Learning session, I relied heavily on Josh Reich’s machine learning script and other blog posts about predictive analytics. There was also a general session on open-source data mining tools. In addition to R, there was an interesting demo of KNIME (a workflow-based data mining tool that reminded me of Insightful Miner). It was interesting to see that KNIME can run algorithms from R and Weka, too.

    All in all, a really invigorating event. Many thanks to the folks from the ACM for organizing it.

    SF Bay ACM: 2009 Silicon Valley Data Mining Camp
49. Ken says:

    April 11, 2010 at 8:50 am

    Here are the slides from my talk
    Recommended Link: the public terabyte dataset project page (http://bixolabs.com/datasets/public-terabyte-dataset-project/).