Internet　economic　markets　have　upgraded　the　traditional　ways　of　product,　service,　consumption,　and　processing　transactions　and　prompted　the　reforming　of　traditional　multisided　markets.　With　this　background,　our　paper　presents　a　multi-agent　deep　reinforce　learning　method　with　group　hybrid　abstraction　called　GHA-DQN　to　build　Internet　marketplace　model　and　observe　the　evolution　process　including　initiation　and　competition　periods　from　the　areas　of　e-transport,　e-commerce,　e-catering,　and　other　e-markets.　The　experiment　results　provide　references　to　support　networking,　financing,　and　competition　equilibrium　in　this　competitive　e-economic　world.　©　2022　IEEE.