There are a few ways to reduce the false positive rate. First, you can ensure you're using the optimal number of hash functions. Check the Wikipedia page on Bloom filters to see what that is. Second, you can increase the size of the Bloom filter. Third, you can use a more space-efficient filter like a ribbon filter.

m – number of elements in bit array n – number of items in collection p – false positive probability // 0.0 – 1.0 ^ – power

p = e^(-(m/n) * (ln(2)^2));

I wrote a math friendly tutorial on Bloom Filters : http://techeffigy.wordpress.com/2014/06/05/bloom-filter-tutorial/