Define a neural network in keras that is at least as good as random forest classifier

NJACKWinterOfCode / Credit-Card-Fraud-Detection

Uses Machine Learning and Deep Learning algorithms for detecting fraud in credit card transactions

0 stars 4 forks source link

Define a neural network in keras that is at least as good as random forest classifier #9

Open confifu opened 4 years ago

confifu commented 4 years ago

Random forest classifier gave us the following result for the fraud case:

precision : 0.73
recall : 0.97
f1 score: 0.84
Our neural network should perform at least as good as this. This result is already quite good for real world. We predicted 97% of the fraud cases. But also 27% of predicted fraud cases were actually not fraud. We should aim to increase the precision but not at the cost of recall because a false negative is better than a false positive in this case.

confifu commented 4 years ago

Please assign me this issue.@Nish-19

Nish-19 commented 4 years ago

@saturn660 and @sakshamj74 you can work on different deep learning frameworks like Keras(or TensorFlow), PyTorch, etc. if you want to both contribute to the same issue. The choice is yours.

sakshamj74 commented 4 years ago

@saturn660 will you work on pytorch??

sakshamj74 commented 4 years ago

@Nish-19 Please review my pr #10

SubhradeepSS commented 4 years ago

@Nish-19 can I make a tensorflow model?

Nish-19 commented 4 years ago

@SubhradeepSS, yes you can.