Hello! Everyone
I have thousands of text documents which I need to split into several known categories. It's possible that a document matches several of the categories. I am now using the ant colony algorithm for text classification. It is suitable for text classification. I am searching and look for other site that they answered which it's described there as suitable for problems that resemble TSP. Does text classification really resemble TSP? Is it trust?
Which algorithm should I use to do the job? I don't need to do it fast. I need to be sure that the algorithm categorizes correctly (as far as possible). I wanted to use ant colony optimization algorithm. Is there an implementation of in java?
Thank you for your help!
Dear Panei San,
I'm not an expert in text classification, but ACO was originally created for search tasks in problems that can be modeled using graphs (such as TSP). Even though there are many algorithms based on ACO and that can maybe solve such problems, I would look for techniques which were designed to solve classification problems.
Therefore, I think the question you should answer first is: Why should you use ACO for such problem? Do you think you can model your problem in order to make it suitable for this approach?
Best regards,
Marcelo Lacerda
There is also another point I forgot to say. You mention that time is not a constraint to you. Such heuristics are usually used to find the best solution for a given problem without testing all possibilities in the search space. Therefore, they were made in order to solve such problems without consuming too much execution time. However, they have a drawback: they do not guarantee the best solution possible, but stilll a very good one.
If you'd rather prefer a good classification performance instead of a time saving approach, I would recommend for you to consider not to use such techniques, but to chose exact algorithms instead, since they will give the best solution possible, even though facing execution time issues (specially for highly dimensional poblems).
I hope it helps =)
Dear Panei San,
As Marcelo very well stated, ACO is an optimization algorithm and it is used to find the "optimum" solution in a space of possible solutions (configurations). ACO is not a classification algorithm to directly apply to Text classification problem. For text classification, you should consider using well known algorithms like k-NN, Naive Bayes (or variants), etc.
Good luck.
Thanks my dear Marcelo ,
Your advice is very valuable for me. When I read your suggestions, I am more thinking and what methods are more suitable for text classification. Here, I want to know that you say what exact algorithms can be choose for more appropriate. Please can you any suggestion. I am waiting you. Thanks a lot..
Thanks dear Volkan,
your suggestion is to use such as k-NN, Naive Bayes (or variants), etc. It is well known algorithm. It is useful for text classification. I want to really know about how to improve and contribute new features from these well known algorithm. Please can you suggest me. Thanks a lot .
ACO has a broad application in Clustering and extraction of Classification rules. There are a variety of algorithms which you could consider.
sorry i'm not expert in clustering and classification, but i know that Ant-based Clustering methods have shown a good performance.
Dear Panei,
As Mr. Volkan said, these are good examples of classification methods. However, unfortunately, I could not say which ones are more suitable fore such problems, since I'm not an expert on text classification. I would recommend for you to take a look on symbolic AI techniques, such as Classification Trees algorithms (ID3, C4.5, etc). I have a feeling that you should go this way for problems like this one, and not towards connectionist AI. But, of course, I might be wrong.
Best regards,
Marcelo Lacerda
There are some papers related to learning decision rules using ACO; these method could be applied in text classification.
Dear Rafael,
I will see some papers about ACO for text classification from online. I am confused when ACO is more complex than other method. So I think "Should ACO use in text classification" . If it is used, time complexity is higher than other methods. Can you suggest it what method is more suitable for classification.
Thank you.
ant Colony is suitable for solving Graphed problems like irrigation or electrical network.if you want no solve every problem you should change it to a graph.you call me: +98917338620, availible in Viber,Phone
Hello Panei San: see the paper: http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5513964&tag=1
This paper discusses the use of ACO to scale down text features under consideration before a content classification is done for web pages.
Article An Ant Colony Optimization Based Feature Selection for Web P...
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs