Penyakit stroke menduduki peringkat pertama penyebab kecacatan dan kematian peringkat tiga di dunia sesudah penyakit kanker dan jantung. Masyarakat Indonesia banyak yang belum mengetahui dan menyadari tanda-tanda atau gejala stroke sedini mungkin karena gejalanya beragam. Apalagi masyarakat enggan memeriksakan diri ke rumah sakit untuk berkonsultasi terkait gejala yang dialami.rnOleh karena itu, untuk memudahkan dalam mendiagnosa pasien penyakit stroke dilakukan klasifikasi dengan menggunakan metode K-Nearest Neighbor (K-NN). Atribut yang digunakan pada skripsi ini adalah id, gender, age, hypertension, heart_disease, ever_married, work_type, residence_type, avg_glucose_level, bmi, smoking_status, stroke. Dalam mengetahui suatu tingkatan akurasi nilai dari data yang dihasilkan di lanjutkan dengan proses perhitungan recall, akurasi, dan presisi, dan dengan persamaan. Pengujian dilakukan dengan menggunakan nilai tetangga k =3, k=5, k=7, k=9, dan k=11 untuk setiap pembagian data testing dan data training, penelitian mendapatkan nilai evaluasi terbaik dengan nilai presisi 95.59%, recall 100.00%, f-1 score 97.74%, dan akurasi 95.59% dengan pembagian data sebanyak k=5 pada k-folding 5.


Deskripsi Lain

Stroke is the third leading cause of disability and death in the world after cancer and heart disease. Many Indonesian people do not know and are aware of the signs symptoms of a stroke early possible because the symptoms vary. Moreover, people are reluctant to go to the hospital to consult regarding the symptoms they are experiencing. Many approaches have been taken to predict disease by comparing the technology and performance of predictive data mining. Classification is a technique used to create data models that have not been classified. The purpose of this system is to produce a classification system for early diagnosis of stroke using the K-Nearest Neighbor (K-NN) method which can help and facilitate health workers in diagnosing stroke patients. The attributes used in this thesis are id, gender, age, hypertension, heart_disease, ever_married, work_type, residence_type, avg_glucose_level, bmi, smoking_status, stroke. In knowing a level of accuracy of the value of the data generated, it is continued with the process of calculating recall, accuracy, and precision, and with equations. Tests were carried out using neighboring values k 3, k 5, k 7, k 9, and k 11 for each distribution of testing data and training data, the study obtained the best evaluation value with precision 95.59%, recall 100.00 %, f-1 score 97.74%, and accuracy 95.59% by dividing the data by k 5 in k-folding 5.