Window　opening　behavior　is　closely　related　to　indoor　air　quality,　thermal　comfort,　and　building　energy　saving.　This　study　took　three　typical　rooms　(standard　office,　meeting　room,　and　smoking　office)　of　an　office　building　in　Xi’an　as　a　research　object　in　summer,　and　the　window　state　assessment　and　physical　environment　measurement　were　carried　out.　Three　typical　window　opening　behaviors　were　found,　namely　active,　moderate　and　passive.　The　change　trend　of　the　window　opening　probability　was　observed　to　be　＇twin　peak＇　and　＇inverted　U＇.　This　paper　used　the　binary　logistic　regression　model　to　find　that　the　solar　radiation　has　the　greatest　influence　on　the　window　opening　behavior　for　the　smoking　office　and　standard　office　in　summer.　However,　the　state　of　window　in　the　meeting　room　had　the　strongest　correlation　with　the　time　of　day.　This　study　discussed　the　dividing　principles　involved　in　setting　the　dummy　variable　interval　level　and　adopted　a　combined　statistical　method　to　determine　the　optimal　interval　level　combination　of　each　dummy　variable.　This　study　provided　theoretical　support　for　applications　to　simulate　energy　consumption　in　office　building.　©　2023　IBPSA.All　rights　reserved.