how to interpret the case when the cross validation accuracy is more than the model accuracy I've trained a ANN model which resulted in 94.62%, but when I do a 5 fold cross validation the mean accuracy is 94.75%. Also 4 out of 5 cross validated models accuracy is more than 94.62%.
How to interpret this?
Is this due to possible duplicate data?
or overfitting of the model?
(The ANN models is trained for a binary classification problem and by accuracy I mean balanced accuracy)
 A: Assuming validation and training sets are different as they should be, overfitting is considered when the reverse happens, i.e. when training performance is better than validation performance. There is no reason to think about overfitting in this case. Also, your results are close to each other. Informally speaking, the difference doesn't seem to be statistically significant. Having similar training and validation success is in general good, except underfitting. However, it can probably be better, but I wouldn't consider ~95 % balanced accuracy as underfit.
If validation set consist of samples that are also present in the training set due the duplicate data as you've noted, it's normal that you obtain slightly better results. Some portion of the test data might be luckily sampled from a region where the distribution is well-represented, where the other portion is duplicates from the training part. It all depends on the data quality. 
The entire set of conclusions on the topic depend on the fact that training and test are sampled from same distribution but they're also different. 
