I'm using german credit data which contains 7 numerical and 13 categorical variables for classifying as good or bad applicant. Can I use all the variables for that or I have to consider only the numerical variables for any classification model? However, I need to know about how to use these multiple independent variables in a classification model? plz..
Look at my work. First, you need to perform scale metrization. After that, they will all be viewed from the same point of view. Then, after creating the models, you need to remove poorly identified examples of observations from the source data, because they are not described by Your data. If you give me some of your data I'll show you how to do it
To metrize scales, I use information theory. This is the most common option of the scale. I metrize not only text (categorical) but also numeric scales to handle them all in the model comparably. Standardizing or scaling in another way to 0-1 is an unfortunate decision, because it leads to an irreversible loss of important information.
for categorical variable please use LaberEncoder from sklearn framework and use it for classification.
Link : https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.LabelEncoder.html
you can use the cateogrical data as well. but you have to be carefull when using them. here is main criteria for using categorical data:
whether your categorical data can be ordered by magnitude?
if yes, you can numerical encoding ex: young, mature and old ===> 1,2,3
or not you can use one hot encoding or convert categories to histogram values of them
Article METRIZATION OF MEASURING SCALES OF DIFFERENT TYPES AND JOINT...
To 13 categorical variables you must scale them first , then apply either normalization or standardization to all with the others 7 numerical, before flowed them to you model.
You need to map the categorical variables to numerical variables. I recommend the one-hot encoding for this task. And later you can use a wrapping model to select the best variables for use. Once you get the best variables you can use logistic regression or K-NN to make a classification problem. You can evaluate your model with a confusion matrix or with ROC curves.
Um bom começo é transformar sua variáveis categorias em numérica, após essa etapa veja o que vai eleger como alvo e utilize o algorítimo de classificação mais adequado.
You can scale (mapping)the 13 categorical variables, then normalized all variables (7and the scaled 13), before feeding the classifier on them.
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification