import pandas as pd
from scipy.sparse import csr_matrix

#reading the csv-file
#if yor csv-file has header then header=1
df = pd.read_csv("filepath here", header=1)

#if no header names:
df.columns = ["user_id", "group_id", "group_value"]

#handling missing values in group_value
df["group_value"].fillna(df["group_value"].mean)

matrix = csr_matrix(df)