First of all, composition of functions is always associative (in a way this is "the mother of all associativities"), because for all $x$
$$ ((f\circ g)\circ h)(x)=(f\circ g)(h(x))=f(g(h(x)))=f((g\circ h)(x))=(f\circ(g\circ h))(x).$$
Thus, what you need to show about your set $X$ of functions is
- if $f,g\in X$, then $f\circ g$ in $X$.
- there is a neutral element (which is $f_1$, obviously)
- for each $f\in X$ there is an inverse element, i..e some $g$ with $g\circ f=f\circ g=f_1$
One can also try this the other way around:
The set of bijections $\mathbb Q\setminus \{0,1\}\to\mathbb Q\setminus \{0,1\}$ is certainly a group $G$.
If we start with $f_2,f_3\in G$, there is certainly a smallest subgroup $\langle f_2,f_3\rangle$ of $G$ that contains both $f_2$ and $f_3$.
To find this group (noting that $f_2, f_3$ are their own inverses), we can simply keep playing with composing found functions with $f_2$ and $f_3$ until we no longer find new functions - a process that fortunately terminates and gives us exactly the set $X$.
Another interesting point to note is that all elements of $X$ have the form
$f(x)=\frac{ax+b}{cx+d}$ with $ad-bc\ne0$ (where we ignore the fact that $f$ is not defined at $x-\frac dc$, say we view these as functions on $\mathbb R\setminus \mathbb Q$). This reminds us of invertible matrices $A=\begin{pmatrix}a&b\\c&d\end{pmatrix}$ with $\det A=ad-bc\ne0$. Do you see how the composition of two fucntions of this "matrix type" relates to a well-known operation among the corrsponding matrices?