Classification when my column data is string ID           CAUSABAS   CAUSAIA    MP
2740.0       C509       C509       2
2770.0       C64X       C64X       2
17029.0      C921       C921       1
5015.0       C910       C504       1
5018.0       C504       C539       2
5030.0       C349       C509       2

This is a brief look at my data. The data has more columns. The CAUSABAS   CAUSAIA columns are labels, but it is a string. What kind of techniques in python I can utilize string for classifications? 
I checked on online sources as well. 
Can I use CountVectorizer and token_pattern=TOKENS_ALPHANUMERIC to deal with this problems? 
 A: Classification can be understood as a linear model.
$dep. variable = \beta * ind. variables + \epsilon$

Dependent variable is a string
When the dependent variable is categorical it is a classification. When the dependent variable is numeric it is a regression. A strings can be understood as factor/ unordered categorical variable. Therefore you can use classification algorithms like Multinomial logit. If the strings are ordered categories you should sticked to an algorithm of ordered classification as ordered logit. 
One of the independent variable is a string
Include dummy variables for each category. This only works if you have a small amount of categories. Take on category, e.g. C509, as base category and add a dummy variable for each of the other categories (C64X, C921, C910, and so on). 
If both the dependent and the independent variable is a string you should do both: convert the independent variables to dummy variables and the dependent variables to categorical variables.
To sum it all up
Convert the string to a factor (code in python) and continue like in other classification processes.
Additional hint: multicollinearity
In the first 3 of the 6 observations of your sample CAUSABAS and CUSAIA are the same. If the two variables are the same for many observations you might have problems with multicollinearity. Most Classification algorithms are not working properly if you have multicollinearity.  You can use the VIF (Variance Inflation Factor) to check for multicollinearity. 
