Bug in Class DataDAO about librec

[guoguibing]

The reason why case this bug is that the columns number in trust.text is not start from 1,when the columns number start from 1 no bug found!

What does it mean? I didn't get you. In LibRec, the input ids will be mapped as the inner ids, so they are different.

[Jakye]

I know input ids will be mapped as the inner ids.If the input rating record is(1,1,5),the inner mapping record is (0,0,5).However what I got by method readData(int[] cols, double binThold) of class DataDAO was not mapping inner ids.
If there is just one trust record between users in the original trust.txt ,naming (1,2,1) meaning that user 1 trusts user 2,then according to the following codes
// inner id starting from 0
int row = userIds.containsKey(user) ? userIds.get(user) : userIds.size();
userIds.put(user, row);
int col = itemIds.containsKey(item) ? itemIds.get(item) : itemIds.size();
itemIds.put(item, col);
dataTable.put(row, col, rate);
colMap.put(col, row);
in method readData of Class DataDAO.
row =0, userIds(1,0) ,col=0,itemIds(2,0) ,dataTable(0,0,1),colMap(0,0)
The data is dataTable meaning the inner mapping trust record between user 1 and user 2 is 1 .
Actually ,the correct the inner mapping data in dataTable is (0,1,1) meaning the inner trust record between user 1 and user 2 is 1 ..
The following is my test example data in trust.txt
1 2 1
1 3 1
1 6 1
1 8 1
2 3 1
2 4 1
2 6 1
3 1 1
3 7 1
4 2 1
4 5 1
4 6 1
5 2 1
6 7 1
7 3 1
8 1 1
8 7 1
According to your original idea in method readData of Class DataDAO .The inner mapping trust relationships between users are :
0 1 1
0 2 1
0 5 1
0 7 1
1 2 1
1 3 1
1 5 1
2 0 1
2 6 1
3 1 1
3 4 1
3 5 1
4 1 1
5 6 1
6 2 1
7 0 1
7 6 1
But What I got are:
1 1 1.000000
1 2 1.000000
1 3 1.000000
1 4 1.000000
2 2 1.000000
2 3 1.000000
2 5 1.000000
3 6 1.000000
3 7 1.000000
4 1 1.000000
4 3 1.000000
4 8 1.000000
5 1 1.000000
6 7 1.000000
7 2 1.000000
8 6 1.000000
8 7 1.000000

[guoguibing]

Did you figure out the issue?

[Jakye]

Not yet,I try to fix it!

[guoguibing]

The issue is that the DataDAO does not know both rows and columns are users. By default, it will treat them separately. For this purpose, you need to load ratings first, so that you can build a userIds map first, and then use it as an argument to build a DataDAO object with the following constructor:

public DataDAO(String path, BiMap<String, Integer> userIds, BiMap<String, Integer> itemIds)

[guoguibing]

You'd better read through the source code of SocialRecommender to figure it out.

[Jakye]

I create a test Class which calls the method readData(int[] cols, double binThold) of Class DataDAO to load trust values in social network.I loaded the trust information like rating information.User ids in row were treated as user id ,User ids in column were treated as item ids .I take another case as an example.
If there is just one rating record between user 1 and item 2 in the original rating.txt ,naming (1,2,3.5) meaning that the rating of user 1 for item 2 is 3.5,then according to the following codes
// inner id starting from 0
int row = userIds.containsKey(user) ? userIds.get(user) : userIds.size();
userIds.put(user, row);
int col = itemIds.containsKey(item) ? itemIds.get(item) : itemIds.size();
itemIds.put(item, col);
dataTable.put(row, col, rate);
colMap.put(col, row);
in method readData of Class DataDAO.
row =0, userIds(1,0) ,col=0,itemIds(2,0) ,dataTable(0,0,3.5),colMap(0,0)
The data is dataTable meaning the inner mapping rating record between user 1 and item 2 is 3.5 .
Actually ,the correct the inner mapping data in dataTable is (0,1,3.5) meaning the inner rating record between user 1 and item 2 is 3.5 .
dataTable(0,0,3.5) means the record between user 1 and item 1 is 3.5

[guoguibing]

The input user id "1" is mapped to inner user id 0, while the input item id "2" is mapped to inner item id 0. The class works correctly. I didn't see why it should be (0,1, 3.5).

[guoguibing]

You can use

String user = rateDao.getUserId(u);
String item = rateDao.getItemId(j);

to recover the original user or item ids.

[Jakye]

The method readData(int[] cols, double binThold) of Class DataDAO calls Constructor method of SparseMatrix
// build rating matrix
rateMatrix = new SparseMatrix(numRows, numCols, dataTable, colMap);
This constructor method actually calls the construct(Table<Integer, Integer, Double> dataTable, Multimap<Integer, Integer> columnStructure) method in Class SparseMatrix.

// set data
for (Cell<Integer, Integer, Double> en : dataTable.cellSet()) {
int row = en.getRowKey();
int col = en.getColumnKey();
double val = en.getValue();

        set(row, col, val);
    }

If the parameter of dataTable in method construct is(0,0,3.5) ,then set (0,0,3.5) meaning the record between user 1 and item 1 is 3.5

[guoguibing]

This conversation is over.

[Jakye]

You will encounter this bug sooner or later !

[Jakye]

I think the code in readData(int[] cols, double binThold) of Class DataDAO should be the following
// inner id starting from 0
int row = userIds.containsKey(user) ? userIds.get(user) :(Integer.parseInt(user)-1);
userIds.put(user, row);
int col = itemIds.containsKey(item) ? itemIds.get(item) :(Integer.parseInt(item)-1);
itemIds.put(item, col);
instead of
// inner id starting from 0
int row = userIds.containsKey(user) ? userIds.get(user) : userIds.size();
userIds.put(user, row);

int col = itemIds.containsKey(item) ? itemIds.get(item) : itemIds.size();
itemIds.put(item, col);