I managed to do it using both pgfplot and tikz-datavisualization. Though pgfplot solution is better.
Pgfplot solution
% create a new table with 10 rows and columns 'x' and 'y':
\pgfplotstablenew
[
% define how the 'new' column shall be filled:
create on use/x/.style ={ create col/expr ={100*rand}},
create on use/y/.style ={ create col/expr ={100*rand}},
columns={x,y}
] {1000} \mydata
% show it:
%\pgfplotstabletypeset\mydata
\begin{tikzpicture}[scale=1
, every node/.style={}
]
\begin{axis}[
xlabel=$||x|-|y||$,ylabel=$|x+y|$
]
\addplot[ only marks, color = blue ] table [
, x expr=abs(abs(\thisrowno{0}) - abs(\thisrowno{1}))
, y expr=abs(\thisrowno{0}+\thisrowno{1})
] {\mydata};
\end{axis}
\end{tikzpicture}

tikz-datavisualization solution
Almost same but could not get random sampling of interval.
\begin{tikzpicture}[scale=1 , every node/.style={} ]
\datavisualization[
scientific axes=clean, visualize as scatter
, scatter={
style={mark=*
, mark options={color=blue,mark size=2pt} }
}
]
data[ format=function ] {
var i : interval [-100:100];
var j : interval [-100:100];
func y = abs(\value i + \value j) ;
func x = abs( abs(\value i) - abs(\value j) );
};
\end{tikzpicture}

\addplotwith keysx expr,y expr? – percusse Sep 22 '16 at 06:02datavisualizationlibrary usingfunc xandfunc y. I'll try now. – Dilawar Sep 22 '16 at 06:05