Fig. 5. Histogram of the values of

Fig. 5. Histogram of the values of ~~ divided between positive and negative for the range [-100; 100] when the bias term was removed.
performances, as can be observed for the datasets Engytime and Win gnut in Table IV, that although have found the two clusters cores reasonably accurately, flooding with Single link and Ward has led to different performances, as each of which has approximately 30% of the map as non assigned neurons.
VIII. CONCLUSIONS AND FUTURE WORK
A clustering of the self-organizing map using particle swarm optimization with fitness function set as the CDbw va¬lidity index was presented. The particles of the PSO algorithm contain the coefficients to which each type of measure ~~ is multiplied in a linear combination of all the seven measures de-fined with the subsets of patterns inside hyperspheres centered in each neuron. The system aims to find a stable partition of the map by analysing the trade-off between the length of regions of stability defined by varying the minimum size of the clusters (parameter .) and the mean value of the CDbw in that region, which is done by tuning the parameter 1. The final result of the proposed method is dependent on flooding algorithm used to assign unlabeled neurons to the clusters found when the parameter ~ is low.
Future works will focus on examining other validity indices as the fitness function. As the method aims to find global coefficients ~~ that multiplies all connections ~~, current focus consists of examining local coefficients, so that each pair of neurons has its own set of coefficients. The influence of a feedforward neural network with hidden layers so as to form more complex functions is also being considered, as well as the analysis of the influence of the map size in the performance of the method.

Hasil (Bahasa Indonesia) 1:

Fig. 5. Histogram of the values of ~~ divided between positive and negative for the range [-100; 100] when the bias term was removed.
performances, as can be observed for the datasets Engytime and Win gnut in Table IV, that although have found the two clusters cores reasonably accurately, flooding with Single link and Ward has led to different performances, as each of which has approximately 30% of the map as non assigned neurons.
VIII. CONCLUSIONS AND FUTURE WORK
A clustering of the self-organizing map using particle swarm by name"> optimization with fitness function set as the CDbw va¬lidity index was presented. The particles of the PSO algorithm contain the coefficients to which each type of measure ~~ is multiplied in a linear combination of all the seven measures de-fined with the subsets of patterns inside hyperspheres centered in each neuron. The system aims to find a stable partition of the map by analysing the trade-off between the length of regions of stability defined by varying the minimum size of the clusters (parameter .) and the mean value of the CDbw in that region, which is done by tuning the parameter 1. The final result of the proposed method is dependent on flooding algorithm used to assign unlabeled neurons to the clusters found when the parameter ~ is low.
Future works will focus on examining other validity indices as the fitness function. As the method aims to find global coefficients ~~ that multiplies all connections ~~, current focus consists of examining local coefficients, so that each pair of neurons has its own set of coefficients. The influence of a feedforward neural network with hidden layers so as to form more complex functions is also being considered, as well as the analysis of the influence of the map size in the performance of the method.

Hasil (Bahasa Indonesia) 2:

Gambar. 5. Histogram dari nilai-nilai ~~ dibagi antara positif dan negatif untuk rentang [-100; 100] ketika istilah bias yang telah dihapus.
Pertunjukan, seperti dapat diamati untuk dataset Engytime dan Win gnut pada Tabel IV, bahwa meskipun telah menemukan dua kelompok core cukup akurat, banjir dengan link Single dan Ward telah menyebabkan pertunjukan yang berbeda, seperti yang masing-masing memiliki sekitar 30% dari peta dengan neuron non ditugaskan.
VIII. KESIMPULAN DAN MASA DEPAN KERJA
Sebuah pengelompokan peta mengorganisir diri menggunakan partikel optimasi segerombolan dengan fungsi fitness ditetapkan sebagai indeks CDbw va¬lidity disajikan. Partikel-partikel dari algoritma PSO berisi koefisien yang masing-masing jenis ukuran ~~ dikalikan dalam kombinasi linear dari semua tujuh langkah de-didenda dengan himpunan bagian dari pola dalam hyperspheres berpusat di setiap neuron. Sistem ini bertujuan untuk menemukan partisi stabil peta dengan menganalisis trade-off antara panjang daerah stabilitas didefinisikan dengan memvariasikan ukuran minimum cluster (parameter.) Dan nilai rata-rata dari CDbw di wilayah itu, yang dilakukan dengan tuning parameter 1. Hasil akhir dari metode yang diusulkan tergantung pada algoritma banjir digunakan untuk menetapkan neuron berlabel untuk cluster ditemukan ketika parameter ~ rendah.
karya masa depan akan fokus pada pemeriksaan indeks validitas lainnya sebagai fungsi fitness. Sebagai metode ini bertujuan untuk menemukan koefisien global yang ~~ yang mengalikan semua koneksi ~~, fokus saat ini terdiri dari memeriksa koefisien lokal, sehingga setiap pasangan neuron telah menetapkan sendiri koefisien. Pengaruh jaringan saraf feedforward dengan lapisan tersembunyi sehingga membentuk fungsi yang lebih kompleks juga sedang dipertimbangkan, serta analisis pengaruh ukuran peta dalam kinerja metode.