Can anybody explain me DP algo what is it?

It is easy
At first you should create array of squares[], where ith element equals i*i. This array is key to the desicion)

At start dp[] array has  not improved values(obviously dp[i] = i)
(dp[i] is minimal number of squares that in sum equal to i)
then at every cycle add a new square and
try to improve the previous result of dp.
result in - dp[n]. complexity is O(n*sqrt(n))