2022 Lesson 3, Excel version of Titanic - Machine Learning from Disaster

Hi there,

I followed the lesson to build a 3-layer neural network using Excel. I chose a midpoint number of 0.615 to compare with Predicts - (loss-actual)^2 - to predict whether it Survived or not. This number provides the best accuracy rate.

The accuracy rate is 0.824438202 for the training dataset. But when I apply the same parameters to the test dataset, the accuracy rate drops to 0.67942.

I think the problem may be some empty data in the age field in the test dataset. In the course, we filter out the empty data before normalising the data.

Any idea is welcome.