Advanced Search Techniques for Large Scale Data Analytics Pavel Zezula and Jan Sedmidubsky Masaryk University http://disa.fi.muni.cz ¡Teachers: §prof. Ing. Pavel Zezula, CSc. §RNDr. Jan Sedmidubsky, Ph.D. § ¡Inspiration: §Inspired by the book of people from Stanford: §Jure Leskovec, Anand Rajaraman, Jeff Ullman: Mining of Massive Datasets. Cambridge University Press, 2nd Edition, 476 pages, 2014. §Additional information: http://mmds.org/ § Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 2 ¡Introduction ¡Block 1: §Support for Distributed Processing §Retrieval Evaluation §Clustering §Exercises on topics of Block 1 ¡Block 2: §Finding Frequent Item Sets §Finding Similar Items §Searching in Data Streams §Exercises on topics of Block 2 ¡Block 3: §Link Analysis §Search Applications §Exercises on topics of Block 3 § Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 3 ¡Block 1: §Support for Distributed Processing §Distributed file system §MapReduce, Algorithms using MapReduce §Cost model and performance evaluation §Retrieval Evaluation §Retrieval metrics §Clustering §K-means algorithms §Clustering in non-Euclidean spaces §Clustering for streams and parallelism Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 4 ¡Block 2: §Finding Frequent Item Sets §Handling large datasets in main memory §Counting frequent items in a stream §Finding Similar Items §Applications of near-neighbor search §Shingling of documents §Similarity-preserving summaries of sets §Locality sensitive hashing §Searching in Data Streams §The stream data model §Filtering streams Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 5 ¡Block 3: §Link Analysis §Page Rank §Topic sensitive §Link spam §Search Applications §Advertising on the web §Recommendation systems (collaborative filtering) §Mining social-network graphs § Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 6 ¡Goals: 1)We search to get results 2)We ask to find answers 3)We use filters so that the right staff finds us 4)We browse while wandering and way-finding in restricted space ¡ ¡In reality, we move fluidly between modes of ¡ ask, browse, filter, and search Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 8 ¡Defined by software ¡Buy engine, then figure out what it is good for ¡ ¡It often fails because §It is not easy to use §It is not able to handle needed content types Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 9 ¡85% of all web traffic comes from search engines ¡450+ million searches/day are performed in North America alone ¡70%+ of all searches are done on Google sites ¡ ¡Search is the most popular application ¡(second to E-mail??) 10 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) ¡ ¡60% of searchers NEVER go past 1st page of search results ¡The top three results draw 80% of the attention ¡The first few results inordinately influence query reformulation ¡ Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 11 ¡ ¡When we search, our next actions are reactions to the stimuli of a previous search ¡ ¡What we find is changing what we seek ¡ ¡In any case, search must be: ¡ ¡ fast, simple, and relevant 12 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) ¡Elements of global search: ¡ §Users – goals, psychology, behavior §Interface – interaction, affordances, language §Engine – features, technology, algorithms §Content – indexing structure, metadata §Creators – tools, process, incentives Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 13 ¡Assuming information continually and instantaneously available on the web: 1)We are increasingly handing off the job of remembering to search engines 2)When we need answer, we do not think, we go immediately to a nearest Web connection 3)When we expect information to be easily found again, we do not remember it well 4)Our original memory of facts is changing to a memory of ways to find the facts ¡ Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 14 ¡Search is subjective and also depends on visual and emotional attributes, e.g. shocking, funny, etc. ¡ ¡Browser §not clear end-goal; series of unrelated searches; jump across unrelated topics; expects surprises and random search hints ¡Surfer §moderate clarity of end-goal; exploratory actions at the beginning; e.g. planning a holiday ¡Searcher §very clear about what is searching for; completeness and clarity of results are important ¡ Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 15 Context Content Prevalent strategy Scalability data volume number of users (queries) variety of data types multi-aspect queries high low well established cutting-edge research 16 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) Determinism exact match ► similarity precise ► approximate same answer ► variable fixed infrastr. ► mapping Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 18 19 http://www.isgtw.org/sites/default/files/PopularScience_DataIsPower_0.jpg Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) Data contains value and knowledge Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 20 21 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) ¡Given lots of data ¡Discover patterns and models that are: §Valid: hold on new data with some certainty §Useful: should be possible to act on the item §Unexpected: non-obvious to the system §Understandable: humans should be able to interpret the pattern Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 22 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 23 ¡A risk with “Data mining” is that an analyst can “discover” patterns that are meaningless ¡Statisticians call it Bonferroni’s principle: §Roughly, if you look in more places for interesting patterns than your amount of data will support, you are bound to find crap http://2.bp.blogspot.com/_hc_VAw3YEOk/SdjkJijh7xI/AAAAAAAAAKQ/nuapuWjnbqo/s400/DilbertMiningData2.j pg Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 24 ¡Example: ¡We want to find (unrelated) people who at least twice have stayed at the same hotel on the same day §109 people being tracked §1,000 days §Each person stays in a hotel 1% of time (1 day out of 100) §Hotels hold 100 people (so 105 hotels) §If everyone behaves randomly (i.e., no terrorists) will the data mining detect anything suspicious? ¡Expected number of “suspicious” pairs of people: §250,000 §… too many combinations to check – we need to have some additional evidence to find “suspicious” pairs of people in some more efficient way Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 25 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 26 Scalability Streaming Context Quality Usage ¡Data mining overlaps with: §Databases: Large-scale data, simple queries §Machine learning: Small data, Complex models §CS Theory: (Randomized) Algorithms ¡Different cultures: §To a DB person, data mining is an extreme form of analytic processing – queries that examine large amounts of data §Result is the query answer §To a ML person, data-mining is the inference of models §Result is the parameters of the model ¡In this class we will do both! Machine Learning CS Theory Data Mining Database systems Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 27 ¡This course overlaps with machine learning, statistics, artificial intelligence, databases but more stress on §Scalability (big data) §Algorithms §Computing architectures §Automation for handling large data ¡ 28 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) Machine Learning Statistics Data Mining Database systems ¡We will learn to mine different types of data: §Data is high dimensional §Data is a graph §Data is infinite/never-ending §Data is labeled ¡We will learn to use different models of computation: §MapReduce §Streams and online algorithms §Single machine in-memory ¡ Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 29 ¡We will learn to solve real-world problems: §Recommender systems §Market Basket Analysis §Spam detection §Duplicate document detection ¡We will learn various “tools”: §Linear algebra (SVD, Rec. Sys., Communities) §Optimization (stochastic gradient descent) §Dynamic programming (frequent itemsets) §Hashing (LSH, Bloom filters) ¡ Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 30 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 31 Pavel Zezula, Jan Sedmidubsky. Advanced Search Techniques for Large Scale Data Analytics (PA212) 32 http://1.bp.blogspot.com/-i3DMgAzrd78/Tc5rsrBeTtI/AAAAAAAABrQ/5-iQRRIwkMA/s1600/barbeque.bmp How do you want that data?