andruekonst / weighted-random-survival-forest Goto Github PK

Dear Authors

I wonder if there is a way to calculate the AUC for the weighted RSF?

What would be the code for that?

Dear Andruekonst

I hope all is well. and thank you very much for the great work.
I have managed to apply the WRSF on my data and the results showed that concordance index with weights=0.96 and without weights is 0.958.
I have tried the random survival forest on pysurvival and scikit learn and it showed a concordance index of 0.65

Because there is a significant difference between the results of your algorithm without weights (0.958) and that of pysurvial RSF or scikit learn RSF,I thought to double check with you if there is any signs of data-leak ? are your results showing that on the test dataset?

What is the best environment for the package to work?

Is it Windows?
Can it work on Google colab?

Or best to use Linux?

Can it work on Ubuntu?

How to make it work with a simple binary dataset without time column ?

Dear
I hope all is well.
Thank you very much for the hard work.
I have been trying to run the package on Google colab and on my laptop with Windows environment, however not working and getting the following error:
"The latest version of cvxpy doesn't use sum_entries "

I tried to use older version of cvxpy (0.2.9),
But getting another error " module variables. expressions not found"

I wonder if you can very kindly help me fix this error?

Dear Andruekonst

Thank you very much for the great work.
The package is working now on the example datasets.
However, when I do it on my own dataset, I get the following error on google colab :

i indices: 83
Indices computed: #J = 55953
Indices after subspace selection: #J = 3000
Traceback (most recent call last):
File "test_survival_forests_sample.py", line 139, in
fit_time, ci_weighted, ci_without_weights = test_on(dataset(test_size=test_size),
File "test_survival_forests_sample.py", line 54, in test_on
rf.fit(data_x_numeric, pd_y_values)
File "/content/survival.py", line 327, in fit
self.make_weights(x)
File "/content/survival.py", line 369, in make_weights
left = self.predict_trees_subset_row(q, x.loc[i])
File "/content/survival.py", line 497, in predict_trees_subset_row
subset_prediction = [self.predict_row(self.trees[i], row) for i in range(fst, fst + self.n_trees_in_subset)]
File "/content/survival.py", line 497, in
subset_prediction = [self.predict_row(self.trees[i], row) for i in range(fst, fst + self.n_trees_in_subset)]
File "/content/survival.py", line 478, in predict_row
return fast_implementations.compute_hazard(row[self.time_column],
File "fast_implementations.pyx", line 26, in fast_implementations.compute_hazard
def compute_hazard(float time, np.ndarray[DTYPE_t, ndim=1] count_0, np.ndarray[DTYPE_t, ndim=1] count_1, np.ndarray[DTYPE_t, ndim=1] t, int total):
ValueError: Buffer dtype mismatch, expected 'DTYPE_t' but got 'double'