19th October 2002 - Finally completed the ground work

The past two weeks have been frustrating, since a feedback from Pat really made me clear that my initial approach to researching "The Needs of the Web User" could be improved. The problem of my initial approach was my revealing of segments to divide the users into; in the initial draft I had used a clustering approach which is good at identifying a cluster, but it lacks the fundamentals for describing the cluster afterwards. My solution is to keep a much more clear line from the findings of the conjoint analysis to the final modelling of user behaviour as follow:

First, I apply conjoint analysis using bar charts and tables to identify the stimuli utility for all users in general.

Secondly, I identify three revealed segments based on the preference for either content or usability stimuli, these revealed segments are used to specifically identify their similarities and dissimilarities in terms of stimuli preferences.

Third, having described the revealed segments in terms of their behavioural preferences, I use a Bayesian approach to build a decision tree model that will be able to determine the most likely revealed segment based on users "physical" characteristics. Depending on how deep the decision tree is, I am able to describe the users reveal segments with an accuracy varying from 56% to 64%, bearing in mind of course that the potential for statistical error is higher the deeper the decision tree is since the sample data behind the rules are decreasing with the depth of the tree.

My path in terms of identifying the above process has been through a number of empirical tests with programs I developed. This has been more time consuming than I anticipated; however, on the other hand, I have developed and documented a data mining process that will now be used in our next TARGIT Analysis product, so the time has proven to be worth while, although I now strongly feel that this project should come to an end, so I can move on to my next research project and apply my learning there…

At the moment I am looking to put words to all my notes during the research process, and I hope to have a finalized draft in a few days.

Back to journal


Updated 2002-10-19

 
[Home] - [CALM] - [Blog] - [Work] - [Projects] - [Profile] - [References] - [Links]
 

Search and Site Map are in English language only

 
The official homepage of Morten Middelfart and CALM. Computer Aided Leadership and Management (CALM) is an inspirational speculation on where mankind may be heading in the quest to leverage computer potentials for helping individuals and organizations to self-actualize their symbiotic potentials.

The time frame for this well-informed and provocative speculation on relatively near-term and more distant potentials is clearly within mankind's grasp. Dr. Middelfart argues persuasively that within the next one or two decades, symbiotic links with "intelligent machines" will surely leverage people's potentials, far beyond all human progress to date!

Altogether, a tour de force of well-informed contemporary insights and maturely reasoned speculation; affording possible stepping stones and a creative springboard for what may lie ahead. As has been said: "Man's reach should exceed his grasp; else what's a heaven for?”

Keywords:

morten middelfart, morten, middelfart, morton, doctor, dr, ph.d., author, book, calm, computer, aided, leadership, management, human, potential, energy, decision, fear, strength, weakness, business, intelligence, artificial, decision, individual, organization, computers, leaders, managers, future, targit, targit a/s, intelligent dashboard, intelligent analysis, intelligent reporting, balanced scorecard, business scorecard, data warehouse, data mart