Business & Research Idea: Traffic-driven Real-time Navigation

The idea described in this post was born in 2003. Today, many companies provide map & driving guide services, e.g., google maps, mapqeust, yahoo maps, etc. I think most of the search engine-based portals offer the same (or somehow advanced and customized) service. What I feel from those services is they are static. For example, the driving path is suggested based on the shortest path or highway-biased path. The shortest path or highway can be the fastest path, but not always true. We must consider other factors such as traffic zam, weather (for long driving distance), slope and downhill, etc. In addition, the driving path must also take into consideration about the economic path and/or convinience for driving - for example, the number of rest areas, gas prices over the driving path, local taxes attached, and so forth. For tourists, attractive landscapes are also one of the most important factors to decide the driving path.

I started the idea from using on-line (or 'averaged' historical) traffic information to determine the driving path to reduce the driving time. The shortest path is not the fastest path. (But as described, other factors are also important to different users.)

I thought what if we use simulation to get the shortest, fastest, most economic, most attractive, most convinient, etc driving path. Particularly, some algorithms such as multi-pass simulation and OCBA may help for that problem. Sorry for that, I don't have enough time and space to realize this idea. Hopefully, I wish some of you who are interested in to realize it for research or for commercial.

I'm willing to help you if you want to discuss this idea further.

Business & Research Idea: VSM inspires Recommendation System Design

Recently, I have much concerns on text mining. This helps me know much about text mining techniques, especially, VSM (vector space model). Its long history (about 35 years) makes me hard to improve the original VSM technique. But, VSM gave me another chance to apply it to other applications than text mining. The application in my mind is to design and develop a purchase/buying recommendation system (a.k.a., collaborative filtering) based on the shopping patterns as well as one's preferences.

I thoght VSM is the ideal matrix necessary for capturing customer-product relationship. The row is about customers and the columns is about product. If one customer boughts a particular product in past, the corresponding element must be filled with a number (larger than zero). Otherwise, it still remains zero. We can analyze the matrix using traditional data mining technique or using recent kernel method (i.e., kernel matrix or kernelizing). There are many possible ways to fill each element. A binary model is to assign ones or zeros based on the purchase. A real valued model can have other ways - time, number of purchases, user scores, etc. In addition, grouping is the way to make the approach more efficient and more practical for most applications (e.g., food, cloth). (The original idea is best to discrite items such as book, music, etc).

If you wish to know more about this idea, let me know. I'm willing to discuss it.

Business Idea: Clustering matters to Search.

It is hard for me, and possibly all of you, to deny the search technology, particularly on the Internet, has been dramatically advanced. However, I feel it is not necessary yet. Traditional data mining techniques can help for NeXT Internet search. One of disappointed to Search Engine, including Google and Yahoo, and whatever else, is that the returned results are spaghetti. Not only too much results returned, but also not sorted at all.

Therefore, I suggest to use clustering techniques to improve the visualization of the search results. The clustering may be done in on-line among the returned results or done by off-line among the whole indexed stored in the back-end data center. No matter which approach is used, the clustering technique will help to sort and organize the returned results significantly. The first page after searching may be the same as the current one. But, we can have more options to select the category which we want to find. This approach is somehow to merge technologies (e.g., computing tech, artificial intelligence) with human intelligence. The first page has a panel which contains some categories that separates the topics returned. For example, if we typed 'kernel' in the search field, then the category may include 'kernel machine' in data mining, 'computer kernel' in computer OS/science, etc. By clicking the 'kernel machine', we can only obtain the results associated with 'kernel machine/method', while we get articles about linux kernels by clicking the 'computer kernel'. Each category may has sub-categories, for example, the 'kernel machine' may have 'support vector machine', 'string kernel', etc.

Today's computer wants to be like human. I think Google.com also wishes to achieve this goal. The result pages must be automatically generated by computers themselves, without human intervention. However, the best way to know human's intention is to ask them what they want. Using clustering is not the best way to ask such questions, but that is an efficient and feasible way for that purpose. Until now, AI definitely failed to achieve the original vision and objectives. The merging of human intelligence (HI) with AI may realize the original and revised vision if AI alone cannot.

Readers may further be required knowledge about XML, Kernel method, semantic similarity, web content mining, etc. If anyone wants more ideas about this topic or to contact me, let me know. I'm eager to discuss about this.

Barcelona in today not in history

Barcelona at a glance. History appears today and will be future. The one-day visit may make me longer stay in future.

CEIS Workshop Posted.

The first international workshop on 'Enabling reliable and agile cross-enterprise interoperable systems (CEIS)' with a them 'Moving standards forward for impact in the global industrial environment' will be held at March 26 (Monday), 2007 at Madeira Island, Portugal, as part of the third international conference on interoperability for enterprise software and applications (I-ESA'07). The detailed informaion is available via http://iisl.postech.ac.kr/events/ceis.

Bahn Goes Nou Camp

Finally, Bahn goes the FC Barcelona's stadium Nou Camp at Barcelona Spain. It was a historical and phenominal day for me. The city of Gaudi, one of the most famous architect in the World known for Sagrada Familie and Parc Guel, is also the city of FC Barcelona, which is my favorite football club in the world. Unfortunately, there was no home game during my staying in Barclona but I'm believing in some days I'll be there again and watch the game on my eyes.

Late, BUT, Happy New Year.

Something makes me confused.
BUT,
I'M STILL BELIEVING...