Graph　Neural　Networks　(GNNs)　have　been　widely　used　in　recommendation　systems　due　to　their　effectiveness　in　capturing　intricate　interactions　within　graph　structures.　However,　existing　approaches　employ　the　layer-wise　message-passing　scheme　and　can　hardly　extract　the　higher-order　relationships　in　user　and　item　graphs.　Besides,　these　methods　often　rely　on　single-view　relations　and　have　limited　ability　in　extracting　effective　features　with　comprehensive　information.　To　address　these　problems,　we　propose　a　multi-view　contrastive　learning　approach　based　on　higher-order　graph　neural　networks　named　HMCL　to　incorporate　multi-view　information　with　the　higher-order　relationships.　In　detail,　our　model　utilizes　multi-hop　neighbors　with　meta-paths　to　capture　higher-order　relationships.　We　also　integrate　intra-view　and　cross-view　contrasts　to　construct　a　multi-view　contrastive　learning　pipeline　that　carefully　combines　global　and　local　information　to　improve　the　feature　learning　ability　of　GNNs.　HMCL　effectively　utilizes　similar　groups　of　users　and　items　and　enhances　the　robustness　to　data　sparsity.　Experiments　conducted　on　several　benchmark　datasets　demonstrate　that　HMCL　exhibits　superior　recommendation　performance.　Experimental　analysis　also　indicates　that　HMCL　effectively　mitigates　data　sparsity　and　popularity　bias.　©　2024　The　Authors.