Who do I know the direction of the current flow, when I make nodal analysis?
A priori you do not. You can assume a direction and if the current actually flows in the other direction you will get a negative result.
#2: The problem occurs when I user Ohm's law, it can end up in the wrong results if the direction I assumed is wrong. Any solution to that?
No, this problem cannot occur if you use Ohm's law correctly. Ohm's law does not only state the absolute potential difference between two sides of the resistor, it states which sign the potential difference has. The potential is higher on the side the current comes from.
Perhaps post an example problem with your attempt to solve it so we can see where you are going wrong.
If you do everything consistently then you will never get wrong results, even when you assume a wrong direction. Because of this, I always assume that every current is leaving the node. I know that at least one of those assumptions is wrong, but by being consistent I still come out with the right answer.
There are two ways two write KCL...
Personally I find I make fewer mistakes if I use the latter of the two definitions and write an equation that sums to zero. eg..
A + B + C..... = 0
rather than say
D + E = -F -G
Separate names with a comma.