Estimating　the　percolating　cluster　fraction　is　central　to　many　percolation　models.　For　real　networks,　the　total　size　of　clusters　with　loops　can　be　considered　a　plausible　metric　for　this　fraction.　In　this　paper,　we　develop　a　semi-analytic　algorithm　to　estimate　clusters　with　loops　for　both　site　and　bond　percolation　via　modifying　the　message　passing　algorithm.　We　compared　the　estimates　of　the　original　message　passing　algorithm　and　our　modified　version　with　simulation　results　on　four　real　networks.　Our　findings　suggest　that　our　modified　algorithm　can　achieve　accuracy　for　any　real　network,　provided　that　a　sufficient　number　of　possible　states　following　site　or　bond　occupation　are　selected　and　analyzed　to　calculate　the　final　estimate.