Penelitian ini bertujuan untuk meningkatkan akurasi prediksi path loss (pl) untuk jaringan 4g lte di lingkungan perkotaan tarakan, indonesia. tiga model propagasi konvensional (okumura-hatta, sui, dan ericsson 9999) dievaluasi menggunakan data drive test yang dikumpulkan di sepanjang jalan yos sudarso dengan kecepatan 30 km/jam. hasil drive test menunjukkan bahwa 62,94% kekuatan sinyal (rsrp) di lokasi penelitian tergolong sebagai kategori excellent. namun, untuk kualitas sinyal (rsrq), 51,78% berada pada kategori good. klasifikasi jenis medan berdasarkan model propagasi konvensional menunjukkan bahwa model okumura-hatta mengklasifikasikan 77,78% area sebagai suburban, model sui mengidentifikasi seluruh area sebagai terrain b, dan model ericsson 9999 mengelompokkan seluruh area sebagai urban. evaluasi awal menunjukkan nilai rmse yang tinggi, yaitu 56,64 db, 61,56 db, dan 74,17 db, yang secara signifikan melebihi ambang toleransi sebesar 8 db. untuk mengatasi hal ini, algoritma particle swarm optimization (pso) diterapkan dan menghasilkan peningkatan akurasi yang signifikan pada semua model. di antara ketiganya, model okumura-hatta menunjukkan performa paling konsisten, dengan standar deviasi terendah (3,18 db) dan 77,78% cell id mencapai rmse ≤ 8 db, menjadikannya model yang paling andal untuk digunakan di wilayah tarakan tengah.

The purpose of the research is to increase the precision of path loss (pl) prediction for 4g lte networks in the urban environment of tarakan, indonesia. three conventional propagation models (okumura-hatta, sui, and ericsson 9999) were evaluated using drive test data which was collected along yos sudarso street at a speed of 30 km/hour. the results of the drive test showed that 62.94% of the signal strength (rsrp) at the study site was classified as excellent. however, for signal quality (rsrq), 51.78% were classified as good category. the classification of terrain type based on conventional propagation models showed that the okumura-hatta model classified 77.78% of the area as suburban, the sui model identified the entire area as terrain b, and the ericsson 9999 model classified the entire area as urban. preliminary evaluation yielded high rmse values of 56.64 db, 61.56 db, and 74.17 db, which significantly exceeded the tolerance threshold of 8 db. to address this, the particle swarm optimization (pso) algorithm was applied, resulting in significant accuracy improvements in all models. among the three models, the okumura-hatta model showed the most consistent performance, with the lowest standard deviation (3.18 db) and 77.78% of cell ids achieving rmse ≤ 8 db. these findings substantiate the okumura-hatta model as the most reliable model for utilization in the central tarakan area.