kun222 / trust-prediction-from-user-rating Goto Github PK

All the data comes from Epinions Trust Network Datasets

In this project I mainly focus on the Extended Epinions dataset. In the extended epinonins dataset, it contains user_rating.txt, mc.txt, and rating.txt.

1.user_rating.txt:

Trust is the mechanism by which the user makes a statement that he likes 
the content or the behavior of particular user and would like to see more 
of what the users does in the site. Distrust is the opposite of the trust 
in which the user says that they do want to see lesser of the operations 
performed by that user.

2.mc.txt:

Each article is written by a user.

3.rating.txt:

Ratings are quantified statements made by users regarding the quality of 
a content in the site. Ratings is the basis on which the contents are sorted 
and filtered.

Below is the column desrbition:

MY_ID:

This stores Id of the member who is making the trust/distrust statement

OTHER_ID:

The other ID is the ID of the member being trusted/distrusted

VALUE:

Value = 1 for trust and -1 for distrust

CREATION:

It is the date on which the trust was made

Size: 23.3 Megabytes

Columns: 4

Row: 841372

Below is the column desrbition:

CONTENT_ID:

The object ID of the article.

AUTHOR_ID:

The ID of the user who wrote the article.

SUBJECT_ID:

The ID of the subject that the article is supposed to be about.

Size: 40.3 Megabytes

Columns: 3

Row: 1560145

Below is the column desrbition:

OBJECT_ID:

The object ID is the object that is being rated. The only valid objects at the present time are the content_id of the member_content table. This means that at present this table only stores the ratings on reviews and essays

MEMBER_ID:

Stores the id of the member who is rating the object

RATING:

Stores the 1-5 (1- Not helpful , 2 - Somewhat Helpful, 3 - Helpful 4 - Very Helpful 5- Most Helpful) rating of the object by member.

STATUS:

The display status of the rating. 1 :- means the member has chosen not to show his rating of the object and 0 meaning the member does not mind showing his name beside the rating.

LAST_MODIFIED:

The latest date on which the member modified his rating of the object TYPE If and when we allow more than just content rating to be stored in this table, then this column would store the type of the object being rated.

VERTICAL_ID:

Vertical_id of the review.

Size: 683 Megabytes

Columns: 7

Rows: 13668320

In order to improve the program's running effiency, it is essential to use a well processed input dataset. The original text files separates the rater, type of the item and the producer. The key idear in this processing step is to combine the rater who rated the item i, the reivewer who produced the item i, the tpye of the item.

First of all, we need to clean the dataset for the better calculation accuracy and running effiency.

For user_rating.txt, I deleted the CREATION column.

user_rating.txt after being processed:

For rating.txt I deleted LAST_MODIFIED column and TYPE column, and VERTICAL_ID column.

rating.txt after being processed:

I used Scanner inputstream = new Scanner(file) to take input from the rating.txt, and mc.txt;

item[ ]:

  It stores the the item id from user_rating.txt.

user[ ]:

  It stores the user id from user_rating.txt.

item1[ ]:

  It stores the item id from mc.txt.

type1[ ]:

  It stores the type of the item from mc.txt.

review1[ ]:

  It stores the reviewer id from mc.txt.

Here is code segment of paring the rater with the type of rated item.

 public static long search(long i) {
        
        for (int k = 0; k < 1559803; k++) {
            if (i == item1[k]) {
                
                return type1[k];
                
            }
            
        }
        
        return -1;
        
    }
    
    public static void output() {
        
        try {
            PrintStream out = new PrintStream(new FileOutputStream(
                    "C:\\Users\\s6770\\Desktop\\project1\\sources\\rater.txt"));
            
            long output = 0;
            
            for (int i = 0; i < 13668320; i++) {
                output = search(item[i]);
                out.println(output);
            }
            
            out.close();
            
        } catch (FileNotFoundException e) {
            e.printStackTrace();
        }
        
    }

int i =0;

The basic idear is sending every item id in item[ ] to function Search, This function will return the typed of that item. It output the type into the type.txt file.

Here is code segment of paring the rater with reivewer.

 public static long search(long i) {
        
        for (int k = 0; k < 1559803; k++) {
            if (i == item1[k]) {
                
                return reviewer1[k];
                
            }
            
        }
        
        return -1;
        
    }
    
    public static void output() {
        
        try {
            PrintStream out = new PrintStream(new FileOutputStream(
                    "C:\\Users\\s6770\\Desktop\\project1\\sources\\reivewer.txt"));
            
            long output = 0;
            
            for (int i = 0; i < 13668320; i++) {
                output = search(item[i]);
                out.println(output);
            }
            
            out.close();
            
        } catch (FileNotFoundException e) {
            e.printStackTrace();
        }
        
    }

int i =0;

The basic idear is sending every item id in item[ ] to function Search, This function will return reviewer id. It output the id into the reviewer.txt file.

After we get the two text files we can start combing rating.txt, type.txt and reviewer.txt.

Here is the new rating.txt file：

Before we start sorting, it is important to remove all the noisy data(rows with data missing)

After cleaing the data, we can start sort the rows depends on the column feature.

In rating.txt, I sorted rows depends on the user id from smallest value to largest value.

I also create a copy of rating.txt, and I renamed it with category.txt

In category.txt, I sorted rows depends on the reviewer id from smallest value to largest value.

In user_rating.txt, I sorted rows depend on the MY_ID.