[ABSTRAK Dana pihak ketiga (DPK), yaitu dana yang dihimpun bank yang berasal darimasyarakat, perlu dikelola secara efektif dan efisien dengan mempersiapkanstrategi penempatan dana. Salah satu strategi dalam penempatan dana tersebutadalah menyalurkan kembali kepada masyarakat dalam bentuk pinjaman untukDPK yang diprediksi akan mengendap dalam jangka waktu yang cukup lama danmenyimpan DPK dalam bentuk kas, cadangan, atau investasi jangka pendek untukDPK yang diprediksi tidak akan mengendap dalam jangka waktu yang cukuplama menurut definisi bank. Penelitian ini menggunakan data mining untukmemprediksi porsi DPK yang mengendap dari masing-masing nasabahberdasarkan profil demografi dan transaksinya. Penelitian dibatasi pada produk tabungan, dan data yang digunakan untuk proses data mining adalah data profil nasabah dan data transaksi produk tabungan. Metodologi penelitian ini menggunakan pendekatan CRISP DM. Dan metodedata mining yang digunakan adalah teknik decision tree untuk prediksi, analisaklaster untuk proses diskritisasi label kelas yang akan digunakan dalam klasifikasidan menggunakan analisa RFM (Recency, Frequency, Monetary) untukmenyederhanakan nilai pada atribut-atribut yang terkait dengan transaksitabungan. Metode klasifikasi menggunakan algoritma C4.5 dan analisa klastermenggunakan algoritma k-means dan menggunakan WEKA sebagai data miningtools. Hasil dari penelitian ini adalah model untuk memprediksi porsi danamengendap dari nasabah. Dari hasil evaluasi menggunakan perhitungan sensitivity, spesitivity, dan accuracy menunjukan model yang berhasil dibangun memiliki keakuratan yang cukup baik dalam memprediksi porsi dana mengendap. ABSTRACT Third-party funds (TPF), which is funds raised from the public, need to bemanaged effectively and efficiently by preparing a strategic placements. One ofthe strategies is by distributing loan from TPF that are expected to settle for a longperiod of time and store in the form of cash, reserves, or short-term investmentsfor TPF that are predicted will not settle within long period based on definitionfrom the bank. In this study data mining is used to predict portion of TPF thatwill settle for certain period of each customer based on the demographic profileand transaction history. The scope of this study is only for saving account product,and this study uses the customer profile data and transaction data of savingsproducts for data mining process. The research methodology in this study using the CRISP DM approach. Decision tree classification technique is used for prediction, cluster analysis method is usedfor discretization process of class labels to be used in the classification and useRFM analysis (Recency, Frequency, Monetary) to simplify the value of theattributes associated with the transaction of saving account. C4.5 algorithm isused for classification and cluster analysis using k-means algorithm and WEKA isused as data mining tools. The results of this study is the model that can predictportion of TPF that will settle for a certain period. The model evaluation bysensitivity, spesitivity, and accuracy calculation shows that the model hassuccessfully built a good accuracy in predicting of TPF that are expected to settlefor a long period of time. , Third-party funds (TPF), which is funds raised from the public, need to bemanaged effectively and efficiently by preparing a strategic placements. One ofthe strategies is by distributing loan from TPF that are expected to settle for a longperiod of time and store in the form of cash, reserves, or short-term investmentsfor TPF that are predicted will not settle within long period based on definitionfrom the bank. In this study data mining is used to predict portion of TPF thatwill settle for certain period of each customer based on the demographic profileand transaction history. The scope of this study is only for saving account product,and this study uses the customer profile data and transaction data of savingsproducts for data mining process. The research methodology in this study using the CRISP DM approach. Decision tree classification technique is used for prediction, cluster analysis method is usedfor discretization process of class labels to be used in the classification and useRFM analysis (Recency, Frequency, Monetary) to simplify the value of theattributes associated with the transaction of saving account. C4.5 algorithm isused for classification and cluster analysis using k-means algorithm and WEKA isused as data mining tools. The results of this study is the model that can predictportion of TPF that will settle for a certain period. The model evaluation bysensitivity, spesitivity, and accuracy calculation shows that the model hassuccessfully built a good accuracy in predicting of TPF that are expected to settlefor a long period of time. ] |