Using accumarray is probably better, but you can also use the unique function and a for loop to find the sum and the count.
Edit:
I already gave you some pointers you could have tried. Everyone starts out by trying random things until it works. But the code below shows both methods.
x1 = [ 201901, 201901, 201901, 201902, 201902, 201902, 201903, 201903, 201903];
y1 = [1.52, 1.23, 1.42, 1.46, 2.42, 2.35, 3.24, 3.21, 3.44];
x_method_1=unique(x1);
y_method_1=zeros(size(x_method_1));
for n=1:numel(y_method_1)
%you could also find the count with sum(L) and the sum with sum(y1(L))
L=ismember(x1,x_method_1(n));
y_method_1(n)=mean(y1(L));
end
[x_method_2,~,ic] = unique(x1);
y_method_2=accumarray(ic,y1,[],@mean);
y_method_2=reshape(y_method_2,size(x_method_2));
%confirm the methods are equivalent
assert(max(abs(y_method_2-y_method_1))<2*eps)
