2: Data Warehouse and OLAP Technology for Data Mining
• What is a data warehouse? • A multi-dimensional data model • Data warehouse architecture • Data warehouse implementation • Further development of data cube technology • From data warehousing to data mining


What is Data Warehouse?
• Defined in many different ways, but not rigorously.
– A decision support database that is maintained separately from the organization’s operational database. – Support information processing by providing a solid platform of consolidated, historical data for analysis. – W. H. Inmon — “A data warehouse is a subject-oriented, integrated, time-variant, and nonvolatile collection of data in support of management’s decision-making process.”

• Data warehousing:
– The process of constructing and using data warehouses


Data Warehouse — Key Features (1)
• Subject-Oriented:
– Organized around major subjects:
• E.g., customer, product, sales. • Focusing on the modeling and analysis of data for decision makers, not on daily operations or transaction processing. • Provide a simple and concise view around particular subject issues by excluding useless data in the decision support process.

• Integrated:
– Constructed from multiple, heterogeneous data sources:
• RDBs, flat files, on-line transaction records, …

– Applying data cleaning and data integration techniques.
• Ensure consistency in naming conventions, encoding structures, attribute measures, etc. among different data sources • E.g., Hotel price: currency, tax, breakfast covered, etc.

– When data is moved to the warehouse, it is converted.

Data Warehouse — Key Features (2)
• Time Variant:
– The time horizon for the data warehouse is significantly longer than that of operational systems.
• Current value data ⇔ historical perspective info. (past 5-10 yr).

– Every key structure in the data warehouse:
• Contains an element of time [explicitly or implicitly], • But the key of tuples may or may not contain “time element”.

• Non-Volatile:
– A physically separate store of data transformed (copied) from the operational environment into one semantic data store.
• Operational data update does not occur in data warehouse. • Require no transaction processing, recovery, and concurrency control mechanisms. [⇐ on-line database operations]

– Requires only two operations in data accessing:
• initial loading of data and access of data.

Compared w. Heterogeneous DBMS
• Traditional heterogeneous DB integration:
– Build wrappers/mediators on top of heterogeneous DB.
• E.g., IBM Data Joiner and Informix DataBlade.

– Query-driven approach: [costly]
• When a query is posed to a client site, a meta-dictionary is used to translate the query into queries appropriate for individual heterogeneous sites involved, and the results are integrated into a global answer set • Complex information filtering, compete for resources.

• Data warehouse: update-driven, high performance
– Information from heterogeneous sources is integrated in advance and stored in warehouses for direct query and analysis.

integrated. 6 . consolidated.43] – User and system orientation: customer (query) ⇔ market. • OLAP (on-line analytical processing): – Major task of data warehouse system. – View: current. – Database design: ER + application ⇔ star + subject. – Day-to-day operations: purchasing. registration. – Access patterns: update ⇔ read-only but complex queries. manufacturing. – Data analysis and decision making. payroll.1 p. Table 2. etc. – Data contents: current.Compared w. Operational DBMS • OLTP (on-line transaction processing): – Major task of traditional relational DBMS. accounting. detailed ⇔ historical. inventory. local ⇔ evolutionary. banking. • Distinct features (OLTP ⇔ OLAP): [ref.

summarized. multidimensional integrated. key short.1 p. consolidated ad-hoc lots of scans complex query millions hundreds 100GB-TB query throughput. OLAP OLTP users function DB design data usage access unit of work # records accessed #users DB size metric clerk. flat relational isolated repetitive [ref. simple transaction tens thousands 100MB-GB transaction throughput . Table 2.OLTP vs. IT professional day to day operations application-oriented current. response 7 read/write index/hash on prim.43] OLAP knowledge worker decision support subject-oriented historical. up-to-date detailed.

8 . – data consolidation: DS requires consolidation (aggregation. etc.Why Separate Data Warehouse? • High performance for both systems: [asynchronous] – DBMS ⇒ tuned for OLTP: • access methods. recovery. – data quality: different sources typically use inconsistent data representations. summarization) of data from heterogeneous sources. codes & formats which have to be reconciled. multidimensional view. indexing. concurrency control. consolidation. – Warehouse ⇒ tuned for OLAP: • complex OLAP queries. • Different functions and different data: – missing data: Decision support (DS) requires historical data which operational DBs do not typically maintain.

. such as sales. … – Dimension tables..g.g. year). • E. item (item_name. each for a dimension.From Tables and Spreadsheets to Data Cubes • Data warehouses utilize an n-dimensional data model to view data in the form of a data cube. location. allows data to be modeled and viewed in multiple dimensions. branch. …) and keys to each of the related dimension tables. units_sold. – An n-D base cube is called a base cuboid. brand. week. dollars_sold. – The top most 0-D cuboid. amount_budgeted. – A data cube. month. item. type). [group by] 9 . time (day. is called the apex cuboid. • Dimensions: time. • E. quarter. – Fact table contains numerical measures (say. which holds the highest-level of summarization. supplier. – The lattice of cuboids forms a data cube. • In data warehousing literature.

supplier 3-D cuboids 4-D (base) cuboid time.location.item 4 C2 time. item.supplier item. supplier 10 .supplier 2-D cuboids C34 4 C4 time.supplier item.location location.location time.item.supplier item.location time.Cube: A Lattice of Cuboids C 4 0 all time item location supplier 0-D (apex) cuboid 1-D cuboids C14 time.supplier time.location. location.item.

forming a shape similar to snowflake. therefore called galaxy schema or fact constellation.Conceptual Modeling of Data Warehouses • Modeling data warehouses: dimensions & measures – Star schema: A fact table in the middle connected to a set of dimension tables – Snowflake schema: A refinement of star schema where some dimensional hierarchy is normalized into a set of smaller dimension tables. or Galaxy schema : Multiple fact tables share dimension tables. viewed as a collection of stars. 11 . – Fact constellations.

Example of Star Schema time time_key day day_of_the_week month quarter year item Sales Fact Table time_key item_key branch_key location_key units_sold dollars_sold avg_sales item_key item_name brand type supplier_type branch branch_key branch_name branch_type location location_key street city province_or_street country Measures 12 .

Example of Snowflake Schema time time_key day day_of_the_week month quarter year item Sales Fact Table time_key item_key branch_key location_key units_sold dollars_sold avg_sales location location_key street city_key item_key item_name brand type supplier_key supplier supplier_key supplier_type branch branch_key branch_name branch_type city city_key city province_or_street country Measures 13 .

Example of Fact Constellation time time_key day day_of_the_week month quarter year item Sales Fact Table time_key item_key branch_key location_key units_sold dollars_sold avg_sales item_key item_name brand type supplier_type Shipping Fact Table time_key item_key shipper_key from_location to_location dollars_cost units_shipped branch branch_key branch_name branch_type location location_key street city province_or_street country shipper shipper_key shipper_name location_key shipper_type 14 Measures .

A Data Mining Query Language. DMQL: Language Primitives • Cube Definition (Fact Table): – define cube <cube_name> [<dimension_list>]: <measure_list> • Dimension Definition (Dimension Table): – define dimension <dimension_name> as (<attribute_or_subdimension_list>) • Special Case (Shared Dimension Tables): – First time as “cube definition” – define dimension <dimension_name> as <dimension_name_first_time> in cube <cube_name_first_time> 15 .

branch_name. branch. item. supplier_type) – define dimension branch as (branch_key. item_name. city. brand. year) – define dimension item as (item_key. day.Defining a Star Schema in DMQL • Procedure: – define cube sales_star [time. month. street. type. dollars_sold = sum(sales_in_dollars). avg_sales = avg(sales_in_dollars) – define dimension time as (time_key. country) 16 . quarter. day_of_week. branch_type) – define dimension location as (location_key. location]: units_sold = count(*). province_or_state.

item. year) – define dimension item as (item_key. supplier_type)) – define dimension branch as (branch_key.Defining a Snowflake Schema in DMQL • Procedure: – define cube sales_snowflake [time. branch_name. branch_type) – define dimension location as (location_key. country)) 17 . type. province_or_state. location]: units_sold = count(*). branch. supplier(supplier_key. day. street. quarter. item_name. brand. day_of_week. dollars_sold = sum(sales_in_dollars). month. city(city_key. avg_sales = avg(sales_in_dollars) – define dimension time as (time_key.

branch_type) location as (location_key. shipper. avg_sales = avg(sales_in_dollars) – – – – define dimension define dimension define dimension define dimension country) time as (time_key. location]: units_sold = count(*). street.Defining a Fact Constellation in DMQL • define cube sales [time. branch. month. shipper_type) – define dimension from_location as location in cube sales – define dimension to_location as location in cube sales 18 . location as location in cube sales. item_name. type. city. item. from_location. item. day. dollars_sold = sum(sales_in_dollars). province_or_state. day_of_week. unit_shipped = count(*) – define dimension time as time in cube sales – define dimension item as item in cube sales – define dimension shipper as (shipper_key. brand. branch_name. year) item as (item_key. shipper_name. to_location]: dollar_cost = sum(cost_in_dollars). quarter. supplier_type) branch as (branch_key. • define cube shipping [time.

⇐ derived • Holistic: if there is no constant bound on the storage size needed to describe a sub-aggregate... – E. – E. sum(). avg().g. min_N(). mode(). max(). ⇐ aggregate • Algebraic: if it can be computed by an algebraic function with M arguments (where M is a bounded integer).Measures: Three Categories • Distributive: if the result derived by applying the function to n aggregate values is the same as that derived by applying the function on all the data without partitioning. median(). – E.g.g. min(). ⇐ statistical 19 . count(). each of which is obtained by applying a distributive aggregate function.. standard_deviation(). rank().

L.A Concept Hierarchy: Dimension (location) Higher: more general concepts all region country city office Germany Europe ... Toronto M........ Mexico Spain Frankfurt . Vancouver .. North_America Canada . Chan . Wind 20 Lower: more specific concepts ... all .

10} < inexpensive Year Quarter Month Week Day 21 . week} < year • Set_grouping hierarchy – {1.View of Warehouses and Hierarchies Specification of hierarchies • Schema hierarchy – day < {month < quarter..

Dimensions: Product. month.Multidimensional Data • Sales volume as a function of product. and region. Location. Time gio n Industry Category Region Country City Office Year Quarter Month Week Day Product Re Product Month Hierarchical summarization paths 22 .

S. U.A Country Canada Mexico sum Pr 23 .S.A Sample Data Cube Date od uc t TV PC VCR sum 1Qtr 2Qtr 3Qtr 4Qtr sum Total annual sales of TV in U.A.

Cuboids Corresponding to the Cube C C C 3 0 all 0-D(apex) cuboid product 3 1 date country 1-D cuboids 3 2 product. date. country 2-D cuboids date. country 3 C3 3-D(base) cuboid 24 .date product.

Browsing a Data Cube • Visualization • OLAP capabilities • Interactive manipulation 25 .

2.59 • Roll up (drill-up): summarize data – by climbing up hierarchy or by dimension reduction. • Pivot (rotate): – reorient the cube. or introducing new dimensions. • Slice and dice: project and select. Fig. • Drill down (roll down): reverse of roll-up – from higher level summary to lower level summary or detailed data. • Other operations – drill across: involving (across) more than one fact table. 3D to series of 2D planes. 26 . – drill through: through the bottom level of the cube to its back-end relational tables (using SQL). visualization.Typical OLAP Operations ref.10 p.

Shipping Method Customer Orders CONTRACTS AIR-EXPRESS TRUCK Time ANNUALY QTRLY DAILY CITY COUNTRY DISTRICT REGION Location Promotion DIVISION Organization 27 Customer ORDER PRODUCT LINE Product PRODUCT ITEM PRODUCT GROUP SALES PERSON .A Star-Net Query Model Each circle (abstract level) is called a footprint.

Design of a Data Warehouse: A Business Analysis Framework • Four views regarding the design of a data warehouse: – Top-down view: • allows selection of the relevant information necessary for the data warehouse. – Data warehouse view: • consists of fact tables and dimension tables. and managed by operational systems. – Business query view: • sees the perspectives of data in the warehouse from the view of end-user. – Data source view: • exposes the information being captured. 28 . stored.

. orders. invoices. • Typical data warehouse design process: Choose – A business process to model. … – The grain (atomic level of data) of the business process.Data Warehouse Design Process • Top-down. – Spiral: rapid generation of increasingly functional systems. • From software engineering point of view: – Waterfall: structured and systematic analysis at each step before proceeding to the next. short turn around time. e. – The dimensions for each fact table record. – Bottom-up: Starts with experiments and prototypes (rapid). 29 .g. – The measure that will populate each fact table record. bottom-up approaches or hybrid: – Top-down: Starts with overall design and planning (mature). quick turn around.

Multi-Tiered Architecture Metadata other Monitor & Integrator OLAP Server sources Operational Extract Transform Load Refresh Data Warehouse Serve Analysis Query Reports Data mining DBs Data Marts Data Sources Data Storage 1 OLAP Engine Front-End Tools 2 3 30 .

– Independent vs. – Only some summary views are materialized. dep. (directly from warehouse) data mart. • Its scope is confined to specific. 31 . selected groups. • Data Mart: – a subset of corporate-wide data that is of value to a specific groups of users. such as marketing data mart.1 Three Data Warehouse Models • Enterprise warehouse: – collects all of the information about subjects spanning the entire organization. • Virtual warehouse: – A set of views over operational databases.

Data Warehouse Development: A Recommended Approach Multi-Tier Data Warehouse Distributed Data Marts Data Mart Data Mart Enterprise Data Warehouse Model refinement Model refinement Define a high-level corporate data model 32 .

[Informix’s Red-brick] 33 . and additional tools and services. – Include optimization of DBMS backend. – greater scalability. • Hybrid OLAP (HOLAP): – User flexibility.2 OLAP Server Architectures • Relational OLAP (ROLAP): – Use relational or extended-relational DBMS to store and manage warehouse data and OLAP middle ware to support missing pieces.g. e. low level: relational. – fast indexing to pre-computed summarized data. • Multidimensional OLAP (MOLAP): – Array-based multidimensional storage engine (sparse matrix techniques).. high-level: array. • Specialized SQL servers: – specialized support for SQL queries over star/snowflake schemas. implementation of aggregation navigation logic.

year) 34 . year) (customer). (product). n customer). city. (city. (date). (product. introduced by Gray et al.Cube Operation • Cube definition and computation in DMQL: – define cube sales[item. item. year. year]: sum(sales_in_dollars) – compute cube sales • Transform it into a SQL-like language (with a new operator cube by.’96): – SELECT item. customer). city. (date. year) (date. customer). (item. item) (city.product). product. (city. year () (item) • Need to compute these Group-By’s: (city) (year) 2 – (date. (). SUM(amount) FROM SALES CUBE BY item. city.

n • Materialization of data cube: – Materialize every (cuboid) (full materialization). if Li = 1. • Based on size.Efficient Data Cube Computation • Data cube can be viewed as a lattice of cuboids: – The bottom-most cuboid is the base cuboid. 35 . – How many cuboids in an n-dimensional cube with Li levels? T = ∏ i =1 ( Li + 1) ⇒ 2n . none (no materialization). – Selection of which cuboids to materialize. – The top-most cuboid (apex) contains only one cell. etc. or some (partial materialization). access frequency. sharing.

– Grouping is performed on some sub-aggregates as a “partial grouping step”. 36 . hashing. – Aggregates may be computed from previously computed aggregates. and grouping operations are applied to the dimension attributes to reorder and cluster related tuples. rather than from the base fact table.Cube Computation: ROLAP-Based Method (1) • Efficient cube computation methods: – ROLAP-based cubing algorithms (Agarwal et al’96) – Array-based cubing algorithm (Zhao et al’97) – Bottom-up computation method (Bayer & Ramarkrishnan’99) • ROLAP-based cubing algorithms: – Sorting.

37 . VLDB’96) – Smallest-parent: computing a cuboid from the smallest cuboid previously computed. – Cache-results: caching results of a cuboid from which other cuboids are computed to reduce disk I/Os. – Share-partitions: sharing the partitioning cost cross multiple cuboids when hash-based algorithms are used. al.Cube Computation: ROLAP-Based Method (2) • This is not in the textbook but in a research paper • Hash/sort based methods (Agarwal et. – Share-sorts: sharing sorting costs cross multiple cuboids when sort-based method is used. – Amortize-scans: computing as many as possible cuboids at the same time to amortize disk reads.

thereby reducing memory access and storage cost. • Compressed sparse array addressing: (chunk_id. offset) • Compute aggregates in “multiway” by visiting cube cells in the order which minimizes the # of times to re-visit each cell.Multi-way Array Aggregation for Cube Computation • Partition arrays into chunks (a small sub-cube fit in memory). C c3 61 62 63 64 c2 45 46 47 48 c1 29 30 31 32 c0 b3 B 13 9 5 1 a0 14 15 16 B b2 b1 b0 2 a1 3 a2 4 a3 60 44 28 56 40 24 52 36 20 ordering What is the best traversing order to do multi-way aggregation? 38 A .

Multi-way Array Aggregation for Cube Computation AC BC C c3 61 62 63 64 c2 45 46 47 48 c1 29 30 31 32 c0 B 13 9 5 1 a0 2 a1 3 a2 4 a3 14 15 16 28 24 20 44 40 36 60 56 52 b3 B b2 b1 b0 A b0c0 of BC AB 39 .

Multi-way Array Aggregation for Cube Computation a0b0c0 ⇒ a0b0. a0c0. b0c0. AC BC C c3 61 62 63 64 c2 45 46 47 48 c1 29 30 31 32 c0 B 13 9 5 1 a0 2 a1 3 a2 4 a3 14 15 16 28 24 20 44 40 36 60 56 52 b3 B b2 b1 b0 A AB A=40 B=400 C=4000 ⇒ AB=16000 ⇒ AC=160000 ⇒ BC=1600000 40 .

Store only partitions with aggregate value > threshold 41 . 75-78) – Idea: keep the smallest plane in the main memory.Multi-Way Array Aggregation for Cube Computation (cont’d) • Method: the planes should be sorted and computed according to their size in ascending order. fetch and compute only one chunk at a time for the largest plane.12 (pp. – See the details of Example 2. “bottom-up computation” and iceberg cube computation methods can be explored. • Limitation of the method: computing well only for a small number of dimensions – If there are a large number of dimensions.

Indexing OLAP Data: Bitmap Index • Index on a particular column • Each value in the column has a bit vector: bit-op is fast • The length of the bit vector: # of records in the base table • The i-th bit is set if the i-th row of the base table has the value for the indexed column • not suitable for high cardinality domains Base table Cust C1 C2 C3 C4 C5 Region Asia Europe Asia America Europe Index on Region Index on Type Type RecIDAsia Europe America RecID Retail Dealer Retail 1 1 0 1 1 0 0 Dealer 2 2 0 1 0 1 0 Dealer 3 3 0 1 1 0 0 Retail 4 1 0 4 0 0 1 5 0 1 0 1 0 Dealer 5 42 .

g. S-id) where R (R-id. 43 .Indexing OLAP Data: Join Indices • Join index: JI(R-id. join index relates the values of the dimensions of a start schema to rows in the fact table. – E. …) S (S-id. fact table: Sales and two dimensions city and product • A join index on city maintains for each distinct city a list of R-IDs of the tuples recording the Sales in the city – Join indices can span multiple dimensions. …) • Traditional indices map the values to a list of record ids – It materializes relational join in JI file and speeds up relational join — a rather costly operation fact table • In data warehouses.

e..Efficient Processing OLAP Queries • Determine which operations should be performed on the available cuboids: – transform drill. etc. into corresponding SQL and/or OLAP operations. dense array structures in MOLAP.g. roll. • Exploring indexing structures and compressed vs. dice = selection + projection • Determine to which materialized cuboids the relevant operations should be applied. 44 .

hierarchies. – Operational meta-data. • warehouse schema. monitoring information (usage statistics. view. archived. charging policies.Metadata Repository • Meta data is the data defining warehouse objects. data mart locations and contents. derived data definitions. 45 . – Data related to system performance. dimensions. currency of data (active. – The mapping from operational environment to the data warehouse. It has the following kinds: – Description of the structure of the warehouse. or purged). ownership of data. view and derived data definitions. audit trails). • data lineage (history of migrated data and transformation path). – Business data. • schema. • business terms & definitions. – The algorithms used for summarization. error reports.

• Data cleaning: – detect errors in the data and rectify them when possible. 46 . check integrity. • Load: – sort. summarize.Data Warehouse Back-End Tools and Utilities • Data extraction: – get data from multiple. and external sources. • Refresh – propagate the updates from the data sources to the warehouse. consolidate. heterogeneous. and build indices and partitions. compute views. • Data transformation: – convert data from legacy or host format to warehouse format.

based on a statistical model.’98) – pre-compute measures indicating exceptions. • Discovery-driven: (Sarawagi et al. guide user in the data analysis. – Exception: significantly different from the value anticipated. at all levels of aggregation. InExp. – Visual cues such as background color are used to reflect the degree of exception of each cell. 47 . huge search space. and PathExp values) can be overlapped with cube construction.Discovery-Driven Exploration of Data Cubes • Hypothesis-driven: exploration by user. – Computation of exception indicator (modeling fitting and computing SelfExp.

Examples: Discovery-Driven Data Cubes more PathExp Drill down high InExp high SelfExp high InExp Drill down 48 .

Complex Aggregation at Multiple Granularities: Multi-Feature Cubes • Multi-feature cubes (Ross. month: R such that R.sales) from purchases Dependent! where year = 1997 grouping variable of group gi cube by item. – Ex. 1998): Compute complex queries involving multiple dependent aggregates at multiple granularities. region. find the min and max shelf life.price = max(price) – Continuing the last example. max(price). month}. region. region. sum(R. and find the fraction of the total sales due to tuples that have min shelf life within the set of all max price tuples. among the max price tuples. and the total sales among all maximum price tuples. month. et al. • select item. find the maximum price in 1997 for each group. 49 . Grouping by all subsets of {item.

– Analytical processing: • multidimensional analysis of data warehouse data. performing classification and prediction. pivoting. tables. slice-dice. constructing analytical models.Data Warehouse Usage • Three kinds of data warehouse applications. – Information processing: • supports querying. and presenting the mining results using visualization tools. and reporting using crosstabs. • supports basic OLAP operations. basic statistical analysis. OLAP focuses on interactive data aggregation tools. drilling. data mining emphasizes more automation. – Data mining: • knowledge discovery from hidden patterns. charts and graphs. deeper analysis. 50 . • supports associations.

service facilities. • integration and swapping of multiple mining functions. dicing. and tasks. • DW contains integrated. Web accessing. consistent. – On-line selection of data mining functions. – Available information processing structure surrounding data warehouses.From On-Line Analytical Processing to On Line Analytical Mining (OLAM) • Why online analytical mining? – High quality of data in data warehouses. pivoting. reporting and OLAP tools. – OLAP-based exploratory data analysis • mining with drilling. • Architecture of OLAM (OLAP Mining): 51 . OLEDB. etc. cleaned data. algorithms. • ODBC.

An OLAM Architecture (Layering) Mining query User GUI API Mining result Layer4 User Interface Layer3 OLAP/OLAM OLAM Engine Data Cube API OLAP Engine Layer2 MDDB Meta Data MDDB Filtering&Integration Databases Database API Data cleaning Filtering Data Warehouse Layer1 Data Repository 52 Data integration .

• OLAP operations: drilling. 53 . & nonvolatile collection of data in support of management’s decision-making process. snowflake schema.Summary • Data warehouse: – A subject-oriented. dicing and pivoting. – Bitmap index and join index implementations. – Multiway array aggregation. time-variant. slicing. • Further development of data cube technology: – Discovery-drive and multi-feature cubes. fact constellations. – A data cube consists of dimensions & measures. • Efficient computation of data cubes: – Partial vs. MOLAP. full vs. HOLAP. • OLAP servers: ROLAP. integrated. – From OLAP to OLAM (on-line analytical mining). no materialization. rolling. • A multi-dimensional model of a data warehouse: – Star schema.

Philadelphia. Agrawal. In http://www. A. June 1999. Gehrke. A. Efficient view maintenance in data warehouses. S. and S. Washington. Sept. 1997 Int. Gray.References (I) • S. K. and T. 1998 ACM SIGMOD.0. Raghavan. Singh. Chaudhuri and U.htm. MDAPI specification version 2. 232-243. Agrawal. 26:65-74. E. OLAP council. D. 1997. Pirahesh. Abbadi. Sarawagi. Pellow. Deshpande. In Proc. PA. Sarawagi. and S. 54 • • • • • • • . A. 1999 ACM-SIGMOD. D. 359-370. R. Modeling multidimensional databases. Chaudhuri. R. Bottom-Up Computation of Sparse and Iceberg CUBEs. ACM SIGMOD Record. J. A. S. Automatic subspace clustering of high dimensional data for data mining applications. Layman. F. Ramakrishnan. Bombay. Birmingham. Conf. Naughton. 506-521. R. Reichart. J. A. England. In Proc. 1998. Arizona. F. 1997 ACM-SIGMOD. cross-tab and sub-totals. Bosworth. Gupta. Ramakrishnan. 1:29-54. M. June 1998. India. Agrawal. 94-105. M. P. Gupta. Agrawal. Data Engineering. Dayal. Yurek. 1996. 417-427. Venkatrao. Data Mining and Knowledge Discovery. 1996 VLDB. D. An overview of data warehousing and OLAP technology. A. In Proc. and H. J. 1997. April ‘97. Agarwal. Data cube: A relational aggregation operator generalizing group-by.olapcouncil. Seattle. Beyer and R. In Proc. May 1997. R. On the computation of multidimensional aggregates. and P. In Proc.

D. March 1998. Management of Data. and N. Athens. Valencia. John Wiley & Sons. May 1997. Conf. In Proc. Ross. and J. pages 205-216. OLEDB for OLAP programmer's reference version 1. 263-277. Rajaraman. and D. A. of Extending Database Technology (EDBT'98). Conf. Int. Management of Data. Microsoft. Agrawal. 1997. Chatziantoniou. Zhao. March 1998. P. R. Deshpande. Harinarayan. Montreal. Tucson. 1997 ACM-SIGMOD Int. 1998. Thomsen. Srivastava. In Proc. E. In Proc. Complex aggregation at multiple granularities. Conf. Aug. 55 • • • • • • . M. Valencia. June 1996. 1996 ACM-SIGMOD Int. Megiddo. In http://www. In Proc. 116-125. 159-170. Conf. 1997 Int. In Proc. Spain. Int. Srivastava. Ullman. pages 168-182. 1997. of Extending Database Technology (EDBT'98). Greece. Naughton. and J. A. Discovery-driven exploration of OLAP data cubes. K. Conf. Canada. Fast computation of sparse datacubes. Ross and D. An array-based algorithm for simultaneous multidimensional aggregates. Y. Very Large Data Bases.References (II) • V.0. K. OLAP Solutions: Building Multidimensional Information Systems. Sarawagi. Implementing data cubes efficiently. F. D. Arizona.