October 28, 2007 at 10:16 pm
Comments posted to this topic are about the item Mining or Profiling
October 29, 2007 at 2:43 am
...and that assumes that the analysis software is competently done in the first place, generally a rash assumption to make in the absence of long and hard validation.
October 29, 2007 at 3:07 am
Data Mining is a method of searching data with mathematical algorithms to identify trends and hidden patterns, profile customers, and then predict how these trends or customers will behave in the future.
Sub-topic definition: Customer profiling involves searching the data collected from a company's existing customers for patterns that will allow that company to make predictions about who its potential customers are and how those customers will behave.
Data Mining and Profiling through Email Campaigns is one of the many types of aggregation tools we use to cross tabulate and compare products and services viewed online. We offer customized reporting configurations based on your department, customers or product requirements. Our data can be imported into any CRM (Customer Resource Management) program of your choosing by exporting our data.
Lead Generation: It's all about conversion. Gathering as much opt-in data from online forms, ecommerce sales or email campaigns is the fastest way of keeping in touch with your customer base.
October 29, 2007 at 3:11 am
Is that an editorial contribution? Or an advertising message?
October 29, 2007 at 6:23 am
Steve,
I think that data quality/precision is at least as important as data quantity. Especially in scenarios like weather prediction and terrorist determination. At some point, more data isn't improving your forecast by more than the error factor anyway.
--
JimFive
October 29, 2007 at 7:27 am
Plans like this are irresistable to politicians. They are also incredibly corrosive to any free society. Not all things can be reliably predictied, especially when the behavior is not consistent or well understood. such an error in a marketing campaign is pretty harmless, such an error in in law enforcement is very harmful.
We are moving more and more toward a concept of thoughtcrime. Trying to identify people who have not commited a crime, but 'might'. There is generally no 'predictor' of 'terrorist' behavior, because 'terrorists' really have very much in common with millions of other people. Since 'terrorists' are so extremely rare in the population virtually all leads are likely to be wrong, but even investigating them can become a deadly invasion of privacy (what ever happend to 'probable cause' ,anyhow?)
The only possible result of this is that thousands of people will be investigated, possibly have their lives disrupted and few if any 'terrorists' will be stopped. (of course, as has happened in the past, there will be security press conference theater, a few arrests will be made about people planning exotic crimes, but when it actually comes down to a trial, suddenly in the back pages of the newspaper, we will find that the case fell apart, or wound up on some trivial conviction.)
...
-- FORTRAN manual for Xerox Computers --
October 29, 2007 at 7:49 am
The more data you have, the better you should be able to predict something.
Why does this sound like Seldon Mathematics to me?
ATBCharles Kincaid
October 29, 2007 at 8:31 am
Charles,
I think that's where Asimov might have gotten the idea, from economic and mathematical theories. It's an interesting idea, and Asimov might have developed the ideal for this area.
We just have to remember that a 90, 80, or 70% probability is just that. It's a good chance, not a fact.
October 29, 2007 at 9:11 am
"The more data you have, the better you should be able to predict something. Or at least that's one of the things that I learned while studying economics. If we could actually gather enough data about someone or some system, we could determine what the most likely outputs of the system will be."
That sounds like classic Newtonian physics. Newton believed that everything in the universe works like a giant clock, and strict cause and effect relationships govern the movements of every atom in the universe. Thus, if one has enough information (the complete history and situation of any given atom, including all of the atoms around it), one could reliably predict the future path of that given atom, and its relationships to all the other atoms around it.
What Newton did not realize was how the universe really works at the quantum level. The universe really is quite unpredictable. The same thing goes for your scenario - even if you have all the data at your fingertips, you will never be able to fully predict anything with absolute certainty. The most you can hope for is a high probability that your predictions will be close. Even then, you have to be willing to accept the fact that something out of your control could happen to completely invalidate your predictions (the so-called "butterfly effect").
As for me, I am glad that the universe is fundamentally unpredictable. If Newton were correct, there would be no real freedom, only dull, predictable, cause and effect relationships. The universe, and everything thing in it, would be wholly determined, and our perception of freedom would be merely that - only a perception, not a reality.
October 29, 2007 at 9:32 am
Steve Jones - Editor (10/29/2007)
Charles,I think that's where Asimov might have gotten the idea, from economic and mathematical theories. It's an interesting idea, and Asimov might have developed the ideal for this area.
It's not the first time, and won't be the last, that Sci-fi has been ahead of the curve. Look at the "Jutice Machines" encountered by Dr. Who, or Gort for that matter.
ATBCharles Kincaid
October 29, 2007 at 9:48 am
Problems I have seen in the processing of the little data collected in these cases is the little data that is collected. In most science you have to have enough data to validate the base as well as the gross sample be large enough to validate what are anomalies and what really is a trend.
To group the actives of a few into a 'pattern' that would indicate the possibility of a terrorist is quite the task. Take for instance the six sites being viewed over a period of one day by 7 people.
1. Strength of materials specifically concrete
2. Major faults in building or other major structures.
3. Explosives used in demolition
4. Concentration of people in metropolitan areas of a state.
5. Traffic patterns in the local and Map of the local area
6. Constructions companies who built buildings in the same area as identified in #5.
Now who are the 7 people who could have viewed all of the 6 sites above?
1. A student working on a model for his architecture class that would be faultless to bombing.
2. A demolitions expert who is about to demolish a building in the area.
3. A Homeland Security Training specialist who is looking to see if a particular building or area would be a good candidate for a training exercise.
4. A person working in an older building downtown who is planning his exit in case his building is hit.
5. A building inspector who is wondering about safety of buildings in his area.
6. An architect who is wondering about how to make a building more secure and how also to get building materials into the area.
7. An author about to write a book about a terrorist event in his or her hometown.
All would have good reason. But in this day when anyone or anything is a terrorist we would have to look at everyone of them if we are 'mining' data.
The online activities are only indications not facts. They are only good for looking for possibilities. And really in most cases there is not enough to go on, but a good way for companies to bilk more money of the Uncle Sam.
Not all gray hairs are Dinosaurs!
October 29, 2007 at 10:09 am
I find it interesting that you use the emotionaly charged term "profiling" in your title but not in your comments. The negative connotations given to the word "profiling" has resulted in all kinds of time wasting procedures by those in charge of security to avoid the accusation of profiling. Profiling is a valid investigation procedure when the overwhelming majority of suspects fit a given profile. This is not an invasion of privacy or picking on someone.
Profiling quite often represents the reality, PC procedures to avoid it represent a fantasy.
October 29, 2007 at 10:19 am
Steve,
Good points and not sure why I didn't write profiling in the editorial. I wasn't really thinking of the idea of profiling as much as the mining and accuracy of predictions.
The problem, IMHO, with profiling in the past is that it wasn't profiling based on anything other than skin color in many cases. That's a problem with the particular profile, not the idea of profiling in general.
October 29, 2007 at 11:23 am
Even with all the advanced data collection available, advanced programs like that of the Mark V by Dr. Daystrom of the Daystrom Institute failed to correctly determine right from wrong because of its inherent instability of the programming from Dr. Daystrom into the Mark V in 2268.
At least, that is what HAL told me...
The best we are able to do to have programs raise flags and then require a human to interpret them.
I don't ever want a computer making a life or death decision about me be it for health or profiling.
October 29, 2007 at 11:29 am
Steve Holle (10/29/2007)
I find it interesting that you use the emotionaly charged term "profiling" in your title but not in your comments. The negative connotations given to the word "profiling" has resulted in all kinds of time wasting procedures by those in charge of security to avoid the accusation of profiling. Profiling is a valid investigation procedure when the overwhelming majority of suspects fit a given profile. This is not an invasion of privacy or picking on someone.Profiling quite often represents the reality, PC procedures to avoid it represent a fantasy.
Profiling is used in different ways. When used after a crime is committed to use facts about the crime to narrow down the search it can be useful*. When no crime has yet been committed (and may not be) there are no facts on which to base the profile, it is absurdly open ended, and indeed can lead investigators in the wrong directions.
*Even for existing crimes it is of limited usefulness... the 'unibomber' was on of many profiles which were wildly different from the eventual actual convicted perpetrator (to the extent that looking for the profile may have actually hindered the investigation)
...
-- FORTRAN manual for Xerox Computers --
Viewing 15 posts - 1 through 15 (of 22 total)
You must be logged in to reply to this topic. Login to reply