Analytics 2.0

Much more than just tagging

Business Intelligence – When information makes the difference

Posted on | October 9, 2006 | Comments Off

Business solution providers and IT vendors, have focused mainly on scalability, automation, and the ability to merge results from relatively simple analytics with opinion from human experts. For example, “e-business applications” in the areas of SCP (supply chain planning), financial analysis and business forecasting, traditionally rely on decision support systems with embedded “data mining”, operations research and OLAP technologies, business intelligence (BI) and reporting tools as well as an easy to use GUI (graphical user interface) and extensible business workflows (e.g., Geoffrion-Krishnan, 2003).

These applications can be custom built by utilizing software tools, or available as prepackaged “e-business application suites” from large vendors like SAP® (thanks Guille for your quote, I will never forget about it), PeopleSoft® and Oracle® as well as “best of breed” and specialized applications from smaller vendors like Seibel® and i2®.

In this contex, DMT and DSS can help in the discovery of novel patterns, development of predictive or descriptive models, built environments in a sustainable fashion, etc.

Business forecasting, planning and decision support applications (e.g., see Shim et al., 2002; Carlsson and Turban, 2002; Wang and Jain, 2003; Yurkiewicz, 2003) usually need to read data from a variety of sources like on-line transactional processing (OLTP) systems, historical data warehouses and data marts, syndicated data vendors, legacy systems, or public domain sources like the Internet, as well as in the form of real-time or incremental data entry from external or internal collaborators, expert consultants, planners, decision makers and/or executives. Data

from disparate sources are usually mapped to a predefined common data model, and incorporated through extraction, transformation and loading (ETL) tools. End users are provided GUI based access to define application contexts and settings, structure business workflows and planning cycles and format data models for

visualization, judgmental updates or analytical and predictive modeling. The parameters of the embedded data mining models might be preset, calculated dynamically based on data or user inputs, or specified by a power user. The results of the data mining models can be automatically utilized for optimization and recommendation systems and/or can be used to serve as baselines for planners and decision makers. Tools like BI, Reports and OLAP (e.g., Hammer, 2003) are utilized to help planners and decision makers visualize key metrics and predictive modeling results, as well as utilize alert mechanisms and selection tools to manage by exception or by

objectives.

Scientists and engineers have traditionally utilized advanced quantitative approaches for making sense of observations and experimental results, formulating theories and hypotheses, and designing experiments. For users of statistical and numerical approaches in these domains, DMT often seems like the proverbial “old wine in new bottles”. However, innovative use of DMT include the development of algorithms, systems and practices that can not only apply novel methodologies but also scale to large scientific data repositories (e.g., see Han et al., 2002; Connover et al., 2003; Graves, 2003; Ramachandran et al., 2003; He et al., 2003). While scientific and business data mining have a lot in common, the incorporation of domain knowledge is probably more critical in scientific

applications. When appropriately combined with domain specific knowledge about the physics or the data sources/uncertainties, DMT approaches have the potential to revolutionize the processes of scientific discovery, verification and prediction (e.g., Han et al., 2002; Karypis, 2002). This requires the use of DSS, where the

results of DMT can be combined with expert judgment and techniques from simulation, OR and other DSS tools. The “Reviews of Geophysics” (a 1995 publication of the American Geophysical Union) provides a slightly dated discussion on the use of data assimilation, estimation and OR, as well as DSS, (e.g.,http://www.agu.org/journals/rg/rg9504S/contents.html#hydrology).

Business applications have focused on DSS, with embedded and scalable implementations of relatively straightforward DMT. Scientific applications have traditionally focused on advanced DMT in prototype applications with sample data.

Researchers and practitioners of the future need to utilize advanced DMT for business applications and scalable DMT and DSS for scientists and engineers. This provides a perfect opportunity for innovative and multi-disciplinary collaborations.

Terms and Definitions

Analytical Information Technologies (AIT): Information technologies that facilitate tasks

like predictive modeling, data assimilation, planning or decision-making, through

automated data-driven methods, numerical solutions of physical or dynamical

systems, human-computer interaction, or a combination. AIT includes DMT, DSS, BI,

OLAP, GIS, and other supporting tools and technologies.

Business Intelligence (BI): Broad set of tools and technologies that facilitate

management of business knowledge, performance and strategy, through automated analytics

or human-computer interaction.

Business and Scientific Applications: End-user modules which are capable of

utilizing AIT along with domain specific knowledge (e.g., business insights or

constraints, process physics, engineering know-how). Applications can be custom

built or pre-packaged and are often distinguished form other information

technologies by their cognizance of the specific domains for which they are designed.

This can entail the incorporation of domain specific insights or models, as well as

pre-defined information and process flows.

Data Mining Technologies (DMT): Broadly defined, these include all types of

data-dictated analytical tools and technologies that can detect generic and interesting

patterns, scale (or can be made to scale) to large data volumes and help in automated

knowledge discovery or prediction tasks. These include determining associations and

correlations, clustering, classifying and regressing, as well as developing

predictive or forecasting models. The specific tools used can range from

.traditional. or emerging statistics and signal or image processing to machine

learning, artificial intelligence and knowledge discovery from large databases,

as well as econometrics, management science and tools for modeling and

predicting the evolutions of nonlinear dynamical and stochastic systems.

Data Assimilation: Statistical and other automated methods for parameter estimation,

followed by prediction and tracking.

Decision Support Systems (DSS): Broadly defined, these include technologies that

facilitate decision-making. These can embed DMT and utilize these through

automated batch processes and/or user-driven simulations or what-if scenario

planning. The tools for decision support include analytical or automated approaches

like data assimilation and operations research, as well as tools that help the human

experts or decision-makers manage by objectives or by exception like OLAP or GIS.

Geographical Information Systems (GIS): Tools that rely on data management technologies

to manage, process and present geo-spatial data, which in turn can vary with time.

On-Line Analytical Processing (OLAP): Broad set of technologies that facilitate

drill-down or aggregate analyses, as well as presentation, allocation and

consolidation of information along multiple dimensions (e.g., product, location

and time). These technologies are well-suited for management by exceptions or

objectives, as well as automated or judgmental decision-making.

Operations Research (OR): Mathematical and constraint programming, and other techniques

for mathematically or computationally determining optimal solutions for objective

functions in the presence of constraints.

Predictive Modeling: The process through which mathematical or numerical

technologies are utilized to understand or reconstruct past behavior, and predict

expected behavior in the future.

Commonly utilized tools include statistics, data mining and operations research,

as well as numerical or analytical methodologies that rely on domain-knowledge.

Category: Analytics 2.0, Analytics Tips, Web Analytics

Comments

Comments are closed.

Suscribirse al Feed

ESPANOL

.

Analytics 2.0 on Facebook

  • Tag Cloud

    Advanced Implementation Analytics 2.0 Analytics Quote Andres Snitcofsky avinash Basic GA Setup Betazeta blog humor blogs Buenos Aires Web Analytics Day daily quote Email Analytics Email marketing emarketer Endeavor Google Advertisers Google Analytics Google Analytics Configuration interactive vs tradicional iphone iphone users Keepcon market research Martin Enriquez mobile advertising New Indextools New Yahoo Web Analytics online market Popego Quote of the day Red Innova Red Innova Finalists Red Innova Winner Santiago Siri Search Engine Industry Search Geek Setting up Google Analytics Socialmetrix Startups Competition Top Ten Advertisers web analytics day why blogging Y!WA 9.5 Yahoo! Web Analytics Yahoo Web Analytics 9.5

  • Archives

  • Log in / Register

    • .