I have successfully run the movielens examples but there are two questions that I cannot answer yet.
1-Why use timestamps instead of ratings as the features? It seems to me that a rating is a much better feature rather than a timestamp, although I understand that what a user has cheked out recently is relevant, i think that tha rating given to a movie is an even better (at least more intuitive) feature.
2-After i have run the examples, i have tried to predict recomendations for myself, so I created an input file with movies I like and run it, the results however are... almost always the same, example of an input file:
(Sorry, i did not manage that to get properly formatted)
As you can see, the recommendations are almost the same ones.
I have tried this with a number of different examples, and I never seem to be able to get something right, however, if I pass the neuronal network a traning example such as
priyankapatil811
commented
7 years ago
I have successfully run the movielens examples but there are two questions that I cannot answer yet.
1-Why use timestamps instead of ratings as the features? It seems to me that a rating is a much better feature rather than a timestamp, although I understand that what a user has cheked out recently is relevant, i think that tha rating given to a movie is an even better (at least more intuitive) feature.
2-After i have run the examples, i have tried to predict recomendations for myself, so I created an input file with movies I like and run it, the results however are... almost always the same, example of an input file:
` X1 1590,974820748:91231,974820776:
X2 71164,974820691:78316,974820943:7362,974820691:1762,974820943:
X3 5960,974820748:46808,974820889:47446,974820889:26349,974820748:57041,974820889:597,974820748:
X4 58332,974820720:4634,974820943:60338,974820720:79594,974820720:33072,974820776:87572,974820748:7098,974820943:91692,974820748:
X5 86544,974820943:5687,974820776:33679,974820889:48322,974820943:67087,974820691:7214,974820776:84508,974821014:8196,974820943:71748,974820776:7446,974820748: `
The output is
`X1 858,0.922:1221,0.811:2959,0.630:2571,0.597:527,0.555:2023,0.501:6016,0.499:597,0.495:1193,0.484:110,0.481:
X2 858,0.918:1221,0.805:2959,0.623:2571,0.590:527,0.562:6016,0.504:597,0.500:2023,0.498:1193,0.486:110,0.479:
X3 858,0.924:1221,0.811:2959,0.633:2571,0.596:527,0.591:597,0.561:6016,0.516:2023,0.497:500,0.494:1193,0.494:
X4 858,0.919:1221,0.807:2959,0.620:2571,0.593:527,0.564:597,0.503:6016,0.503:2023,0.499:1193,0.487:110,0.479:
X5 858,0.912:1221,0.795:2959,0.628:2571,0.593:527,0.546:6016,0.522:597,0.509:2023,0.487:1193,0.470:4993,0.468:`
(Sorry, i did not manage that to get properly formatted)
As you can see, the recommendations are almost the same ones.
I have tried this with a number of different examples, and I never seem to be able to get something right, however, if I pass the neuronal network a traning example such as
8 1,833981871:3,833981733:6,833982631:10,833981834:19,833981918:21,833981710:39,833981963:47,833981896:48,833982197:110,833981870:150,833973081:153,833973175:161,833981804:165,833973175:168,833982178:172,833982178:173,833981993:185,834586861:207,833982765:208,833981834:231,835087532:236,833982013:253,833981834:266,833981963:276,833982244:277,833982245:288,833981870:292,833981834:296,833973081:316,833973235:329,833973235:338,834586900:339,833981834:344,833973175:349,833973175:350,833982244:353,833982683:355,833982808:356,833982259:357,833982790:364,833982013:367,833982133:372,834587089:377,833982715:380,833973081:381,833982668:434,833981804:442,833982259:454,835087603:457,833981896:480,833982648:500,833982790:508,833982957:527,833973298:539,833982860:550,834587074:551,833982668:552,833982859:553,833982013:587,833982889:588,833973175:589,833982668:590,833973081:592,833973081:593,833981896:595,835087532:597,833982889:610,833981710:648,834586915:733,834586590then i get proper recommendations
8 590,0.831:480,0.821:110,0.770:500,0.763:858,0.758:527,0.719:589,0.718:597,0.712:434,0.686:364,0.651:What am i missing here?