This study propose~s a De~e~p Ne~ural Ne~twork (DNN) frame~work to pre~dict joint re~action force~ ratios in structural analysis using datase~ts obtaine~d from SAP2000 simulations. The~ datase~ts cove~r various load case~s and ge~ome~trical parame~te~rs, e~nsuring the~ mode~l is e~xpose~d to dive~rse~ structural sce~narios. The~ DNN archite~cture~ comprise~s multiple~ fully conne~cte~d laye~rs, e~mploying Re~LU activation functions, dropout re~gularization, and batch normalization for stable~ training. Mode~l pe~rformance~ was e~valuate~d using Me~an Square~d E~rror (MSE~), Me~an Absolute~ E~rror (MAE~), R² score~, and pre~diction accuracy within a 5% e~rror margin critical for civil e~ngine~e~ring applications. The~ re~sults de~monstrate~ e~xce~lle~nt pre~dictive~ capabilitie~s, achie~ving accuracy le~ve~ls e~xce~e~ding 98% across all datase~ts. Notably, the~ third datase~t yie~lde~d the~ lowe~st accuracy at 98.97% and an R² score~ of 0.9915, with slightly e~le~vate~d e~rror me~trics (MSE~ of 5.11, RMSE~ of 2.26, and MAE~ of 1.51). De~spite~ the~se~ challe~nge~s, the~ DNN mode~l consiste~ntly de~live~rs robust pre~dictions, showcasing its pote~ntial for practical structural he~alth monitoring and de~sign optimization. Future~ work should conside~r incorporating more~ dive~rse~ and e~xpe~rime~ntal data to e~nhance~ mode~l robustne~ss furthe~r.