Options

analyzing data that is labeled with positive ( +1) and negative (-1),

charlycharly Member Posts: 6 Contributor I
in Help






Dear Guys ,,

how are you? many thanks for your help ,, i really need your advice ,, firstly , I have analyse my data that is labeled with positive ( +1) and negative (-1), " what i do is I put every single comment in a text file and create 2 folders the first is for positive comments and the other is for negative comments ,then I analyze them i use KNN and SVM " 

Now i have build a lexicon as a CSV file , every word and its weight , i need to enhance my results by using the lexicon weights , so how you can help me to use weights of the new lexicon with the weights that I have from the SVM model ??

what i have to do ?

many thanks in advance , i appreciated your help

Answers

  • Options
    yyhuangyyhuang Administrator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 364 RM Data Scientist
    Hi @charly,
    Do you want to add the weights from SVM as new features? If so, you can join the weight table with the vectorized text. The weights from SVM is usually used to explain the feature importance, select features for dimension reduction, and normalize input data by feature weights. Take a look at the tutorial process of "select by weights" and "scale by weights" operator.
    Cheers,
    YY
  • Options
    charlycharly Member Posts: 6 Contributor I
    hello yyhuang ,,,

    i think to use the update model to update my results and to use the words in lexicon that is matches in the data file that i use before ,, what is ur advice ??

    i will attach my work ,, but the update model tells me there is an error some where ,, i need some help pleease !!!
  • Options
    charlycharly Member Posts: 6 Contributor I
    hello guys ,, i'm confused , i have get to academy and community ,, i didin't find really  what i need !!!

    any help or any informations ??

  • Options
    yyhuangyyhuang Administrator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 364 RM Data Scientist
    edited April 2021
    Hi @charly, thanks for sharing the process for SVM. Do you have input data to run it? I saw you skipped the training step inside cross validation. You directly retrieve a pretrained model and use it for testing. Why? Do you still want to use the feature weights from SVM?
  • Options
    charlycharly Member Posts: 6 Contributor I
    hello @yyhuang  ,thank you , i have an input data that is my whole data , then i have used cross validation  for retrieve data that is use the weights from the SVM , but i want to update weights and enhance them by using the weights from the lexicon ( the CSV file ) , where i can but my new wights ??!! thats my question 
Sign In or Register to comment.