mashimaroljc/ml-and-dm-in-action

Share my code during learning machine learning and data mining

ML and DM in action

中文

##Directory

• Learn-to-identify-similar-images
• BetaMeow
• DouBanMovie
• my-ml-package
• dudulu
• chinese-poem

##Detail

###Learn-to-identify-similar-images

After reading some papers from internet,I implement three algorithms including Average Hash, Discrete cosine transform and so on,which about how to identify the similar images,by Python ,Pillow and OpenCV.

I also complete a program to find the face from picture and identify the simliar face base on the modal called Heer provided by OpenCV.

This is another repository and you can find it following this link.

Learn-to-identify-similar-images

I wrote a article about it to Segmentfault,but I wrote it in chinese.

识别相似图片(一)
识别相似图片(二)

---

###BetaMeow

I create BetaMeow,which can play gobang game with you after learning about the interesting match between AlphaGo and Lee se-dol. Comparing with the previous version,the current version has a great difference from the traditional gobang game AI. It makes a decision by Decision tree instead of search algorithms that the traditional gobang game AI did.

BetaMeow can keep studying only if you provide data. Of course, thought it's not so perfect, I will do my best to update it.

####How To Begin

Download the code and enter following code when you are in the correct directory.

python ai.py

Open your browser and enter http://localhost/five then you can play with BetaMeow

Please make sure that yourPort 80 is free, otherwise you need to modify the code.

####Request

• python >= 3.4
• flask 0.10.1

---

###DouBanMovie

• The web spider made up of spider.py and douban.py baesd on the model named requests. It can collect the information of movie and other video from douban.com，including follow items.

  • title
  • runtime
  • type
  • director
  • actor
  • area
  • star
  • one or more comments

The information will save as JSON format.

You can set the number of information you want to collect.The spider will collect util the queue of mission to be empty when the number is default.Also the spider support breakpoint collecting.The file named info.txt will record the information about the program.**When you continue the process after interruption,this information will be load as configuration.**So, don't modify it unless you need.

• datas.pyis an analysis of the data of the script.At present it can do that.

  • The relationship between actor and runtime.
  • The five actors who has most roles in the good movie or bad movie.You can change the standard to distinguish between good movie and bad movie by change the code.
  • The relationship between area and runtime.
  • The relationship director actor and runtime.
  • The five areas Where is the production of much of the good movie or bad movie.
  • The five directors who has most roles in the good movie or bad movie.
  • The five types which has most good movie or bad movie
  • The relationship between star and runtime.
  • Choose five actors in each type，who has most movie in this type,without good or bad.

• DouBanMovie/data

You can find about 500 randomly selected data about the movie.

• DouBanMovie/result

  You can find the result of running the datas.py based on 500 randomly selected data.

• recommend.py

  Implements an simple algorithm，,which recommend goods based on some labels that customer love recently.It is suitable for a small number of users and individual users.Also,the code is very easy reading and understanding.

• api.py

  This is a web api based on flask and recommend.py.Using it,you can do following things

  • /recommend/get

  Get the information about recommend movie with JSON format.

  • /recommend/put?moviename=anyname&comment=good

  Let the system know Which movie you choose and your comment,like good or bad. Then system will update the model.

####Request

• python >= 3.4
• requests 2.8.1

---

My-ML-algorithm-package

click the link for deatil.

My-ML-algorithm-package

---

dudulu

API server of text mining [dudulu](https://github.com/MashiMaroLjc/dudulu)

---

poem

a chinese poem generator using base rule and probability

To Be Continue……

##Other

The reposiory record the code I write during I learning the machine learnning and data mining. I won't provide a complete data file because they are generally too large to be stored. But I will provide some important examples of data if necessary.

If you are interested in it.you can recommend your project about ML or DM and give me your URL.I will write the URL on my README file .At the same time I also hope you can write URL of this reposiory on README file of your project.

You can open a issue or Pull Request to me if you have a suggestion or better idea.

I will continue update the Repository for a long time,welcome to watch it or give me a star for support，thank you.

##License Apache License 2.0

##Contact Me

Twitter @fatfatrabbit

ml-and-dm-in-action's Issues

能否修改IP

已经解决了，谢谢，不过好像下不了棋

关于五子棋

下子时位置判定有问题，黑子会自动跑到棋盘的一条斜线上下子

默认使用80端口不妥

*nix (Linux, BSD, OS X) 环境下面，小于1024的端口是需要root权限才可以打开的，所以也就是需要sudo才可以运行这个程序，所以不是很妥当
而且，很多开发者都会在本地开着nginx, apache之类的服务器，一般也占着80，所以默认80会造成一些不必要的麻烦

当然了，这个并不是个大问题，可以考虑改个小众端口（也可以设计成从命令行参数读取），供你参考

代码一些问题

试了下代码，python3.6 werkzeug有个bug，导致flask挂掉。
pallets/werkzeug#999
后来转python3.5运行成功。
前端判定棋子位置不是很准确，改了两行

        //var x = parseInt((event.screenX) / 36)-6; 
        //var y= parseInt((event.screenY ) / 36)-2;  
        var x = parseInt((event.offsetX+18) / 36); 
        var y= parseInt((event.offsetY+18 ) / 36);  

Chrome Version 50.0.2661.94 (64-bit) for OS X 没法正常开始游戏

看了下错误控制台并没有任何输出，network所有资源也都加载完成了，有点奇怪

能否提供一下训练的数据集

json

{"neighbor2": {"0": {"neighbor3": {"0": "[-1, -1]", "-1": "[-1, -1]", "1": {"neighbor1": {"0": "[2, 2]", "-1": "[-1, -1]", "1": "[2, 5]"}}}}, "-1": "[-1, -1]", "1": {"neighbor1": {"0": {"neighbor3": {"0": "[3, 1]", "-1": "[-1, -1]", "1": "[1, 5]"}}, "-1": "[-1, -1]", "1": {"neighbor3": {"0": "[3, 5]", "-1": "[-1, -1]"}}}}}}
[2, 5] 的意义是指neighbor2 权重5么？