How to build Mean Squared False Error (MSFE) with Keras Custom loss function

Question

I want to build MSFE loss function.

This function proposed this paper. https://ieeexplore.ieee.org/document/7727770

And here's the concept of it. Please check table, MSE, MFE, and MSFE.

My question is that,

1. Is it possible to divide True Positive, False Postivie, False Negative, True Positive before running model? How could we know y_pred's result even before running prediction model?

In this SO, it is impossible to use this code. Then I think it is impossible. Could you tell me is there any other way?

https://datascience.stackexchange.com/questions/33587/keras-custom-loss-function-as-true-negatives-by-true-negatives-plus-false-posit

TN = np.logical_and(K.eval(y_true) == 0, K.eval(y_pred) == 0)

2. How could I build a custom function for cost-sensitive learning?

Custom loss function in Keras to penalize false negatives Keras Custom loss function to pass arguments other than y_true and y_pred https://datascience.stackexchange.com/questions/28440/custom-conditional-loss-function-in-keras

3. This is what I made. How could I improve this function?

def custom_loss_wrapper(p):
    def custom_loss(y_true, y_pred):
        
        y_true = tf.cast(y_true, tf.float32)
        y_pred = tf.cast(y_pred, tf.float32)

        fp = K.sum(y_pred * (1 - y_true))    
        fn = K.sum((1 - y_pred) * y_true)

        cost = tf.cast(fn * p + fp, tf.float32)
        
        return cost
    return custom_loss