Data Understanding

2.1 Sumber Data

Dataset ECG200 secara historis merupakan bagian dari UCR/UEA Time Series Classification Archive dan saat ini tersedia melalui repositori resmi Time Series Classification.

Asal dan Cara Pengambilan Data

• Asal data: Dataset ini diformat oleh R. Olszewski sebagai bagian dari tesis doktoralnya di Carnegie Mellon University tahun 2001 (Generalized Feature Extraction for Structural Pattern Recognition in Time-Series Data).

• Sumber sinyal: Setiap sampel mewakili satu denyut jantung, direkam dari aktivitas listrik jantung pasien menggunakan sistem perekaman ECG klinis.

• Alat yang digunakan: Rekaman ECG diperoleh menggunakan perangkat elektrokardiografi medis standar, yang terdiri dari:

  • Elektroda permukaan yang ditempelkan pada dada dan anggota tubuh pasien,

  • Amplifier sinyal untuk memperkuat sinyal bioelektrik yang lemah,

  • Analog-to-Digital Converter (ADC) untuk mengubah sinyal analog menjadi data digital berupa deret waktu.

• Proses pengambilan:

  • Sinyal ECG mentah direkam selama sesi klinis,

  • Segmen yang merepresentasikan satu siklus denyut jantung lengkap dipilih secara manual oleh ahli,

  • Segmen tersebut kemudian dipotong, dinormalisasi, dan diberi label berdasarkan diagnosis klinis pasien.

• Jumlah pasien: Dataset tidak berasal dari hanya dua individu. Berdasarkan praktik umum dalam pembuatan dataset ECG medis dan sumber aslinya (termasuk MIT-BIH Arrhythmia Database), 200 sampel kemungkinan besar berasal dari puluhan pasien berbeda. Setiap pasien menyumbang 3–10 denyut jantung yang memenuhi kriteria kualitas dan representasi klinis.

• Waktu pengambilan asli: Data dikumpulkan sebelum atau selama penelitian tesis (sekitar awal 2000-an).

• Format data: Disimpan dalam format ARFF (Attribute-Relation File Format), sesuai standar WEKA.

• Preprocessing oleh penyedia:

  • Setiap sampel mewakili satu denyut jantung, dengan panjang tetap 96 titik waktu.

  • Dinormalisasi secara global.

  • Label: 1 (normal heartbeat) dan -1 (Myocardial Infarction / Ischemia).

Komposisi Dataset

• Data latih: 100 sampel

• Data uji: 100 sampel

• Jumlah fitur per sampel: 96 (time steps)

• Jumlah kelas: 2 (normal vs. infark miokard)

2.2 Eksplorasi Data

2.2.1 Import Library dan Muat Data

import pandas as pd
import numpy as np
from scipy.io import arff
import matplotlib.pyplot as plt
import seaborn as sns

sns.set_style("whitegrid")
sns.set_palette("Set2")

train_path = 'ECG200/ECG200_TRAIN.arff'
test_path = 'ECG200/ECG200_TEST.arff'

data_train, _ = arff.loadarff(train_path)
data_test, _ = arff.loadarff(test_path)

df_train = pd.DataFrame(data_train)
df_test = pd.DataFrame(data_test)

df_train['target'] = df_train['target'].apply(lambda x: int(x.decode('utf-8')))
df_test['target'] = df_test['target'].apply(lambda x: int(x.decode('utf-8')))

print("✅ Data berhasil dimuat dan label telah dikonversi ke integer.")

✅ Data berhasil dimuat dan label telah dikonversi ke integer.

2.2.2 Struktur Kolom dan Informasi Dataset

print("=== Struktur Kolom ===")
print(df_train.columns.tolist())

print("\n=== Informasi Dataset Latih ===")
df_train.info()

print("\n=== Statistik Deskriptif ===")
df_train.describe().T.head(10) 

=== Struktur Kolom ===
['att1', 'att2', 'att3', 'att4', 'att5', 'att6', 'att7', 'att8', 'att9', 'att10', 'att11', 'att12', 'att13', 'att14', 'att15', 'att16', 'att17', 'att18', 'att19', 'att20', 'att21', 'att22', 'att23', 'att24', 'att25', 'att26', 'att27', 'att28', 'att29', 'att30', 'att31', 'att32', 'att33', 'att34', 'att35', 'att36', 'att37', 'att38', 'att39', 'att40', 'att41', 'att42', 'att43', 'att44', 'att45', 'att46', 'att47', 'att48', 'att49', 'att50', 'att51', 'att52', 'att53', 'att54', 'att55', 'att56', 'att57', 'att58', 'att59', 'att60', 'att61', 'att62', 'att63', 'att64', 'att65', 'att66', 'att67', 'att68', 'att69', 'att70', 'att71', 'att72', 'att73', 'att74', 'att75', 'att76', 'att77', 'att78', 'att79', 'att80', 'att81', 'att82', 'att83', 'att84', 'att85', 'att86', 'att87', 'att88', 'att89', 'att90', 'att91', 'att92', 'att93', 'att94', 'att95', 'att96', 'target']

=== Informasi Dataset Latih ===
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 100 entries, 0 to 99
Data columns (total 97 columns):
 #   Column  Non-Null Count  Dtype  
---  ------  --------------  -----  
 0   att1    100 non-null    float64
 1   att2    100 non-null    float64
 2   att3    100 non-null    float64
 3   att4    100 non-null    float64
 4   att5    100 non-null    float64
 5   att6    100 non-null    float64
 6   att7    100 non-null    float64
 7   att8    100 non-null    float64
 8   att9    100 non-null    float64
 9   att10   100 non-null    float64
 10  att11   100 non-null    float64
 11  att12   100 non-null    float64
 12  att13   100 non-null    float64
 13  att14   100 non-null    float64
 14  att15   100 non-null    float64
 15  att16   100 non-null    float64
 16  att17   100 non-null    float64
 17  att18   100 non-null    float64
 18  att19   100 non-null    float64
 19  att20   100 non-null    float64
 20  att21   100 non-null    float64
 21  att22   100 non-null    float64
 22  att23   100 non-null    float64
 23  att24   100 non-null    float64
 24  att25   100 non-null    float64
 25  att26   100 non-null    float64
 26  att27   100 non-null    float64
 27  att28   100 non-null    float64
 28  att29   100 non-null    float64
 29  att30   100 non-null    float64
 30  att31   100 non-null    float64
 31  att32   100 non-null    float64
 32  att33   100 non-null    float64
 33  att34   100 non-null    float64
 34  att35   100 non-null    float64
 35  att36   100 non-null    float64
 36  att37   100 non-null    float64
 37  att38   100 non-null    float64
 38  att39   100 non-null    float64
 39  att40   100 non-null    float64
 40  att41   100 non-null    float64
 41  att42   100 non-null    float64
 42  att43   100 non-null    float64
 43  att44   100 non-null    float64
 44  att45   100 non-null    float64
 45  att46   100 non-null    float64
 46  att47   100 non-null    float64
 47  att48   100 non-null    float64
 48  att49   100 non-null    float64
 49  att50   100 non-null    float64
 50  att51   100 non-null    float64
 51  att52   100 non-null    float64
 52  att53   100 non-null    float64
 53  att54   100 non-null    float64
 54  att55   100 non-null    float64
 55  att56   100 non-null    float64
 56  att57   100 non-null    float64
 57  att58   100 non-null    float64
 58  att59   100 non-null    float64
 59  att60   100 non-null    float64
 60  att61   100 non-null    float64
 61  att62   100 non-null    float64
 62  att63   100 non-null    float64
 63  att64   100 non-null    float64
 64  att65   100 non-null    float64
 65  att66   100 non-null    float64
 66  att67   100 non-null    float64
 67  att68   100 non-null    float64
 68  att69   100 non-null    float64
 69  att70   100 non-null    float64
 70  att71   100 non-null    float64
 71  att72   100 non-null    float64
 72  att73   100 non-null    float64
 73  att74   100 non-null    float64
 74  att75   100 non-null    float64
 75  att76   100 non-null    float64
 76  att77   100 non-null    float64
 77  att78   100 non-null    float64
 78  att79   100 non-null    float64
 79  att80   100 non-null    float64
 80  att81   100 non-null    float64
 81  att82   100 non-null    float64
 82  att83   100 non-null    float64
 83  att84   100 non-null    float64
 84  att85   100 non-null    float64
 85  att86   100 non-null    float64
 86  att87   100 non-null    float64
 87  att88   100 non-null    float64
 88  att89   100 non-null    float64
 89  att90   100 non-null    float64
 90  att91   100 non-null    float64
 91  att92   100 non-null    float64
 92  att93   100 non-null    float64
 93  att94   100 non-null    float64
 94  att95   100 non-null    float64
 95  att96   100 non-null    float64
 96  target  100 non-null    int64  
dtypes: float64(96), int64(1)
memory usage: 75.9 KB

=== Statistik Deskriptif ===

	count	mean	std	min	25%	50%	75%	max
att1	100.0	0.708438	0.593513	-0.706305	0.318643	0.581008	0.994760	2.689017
att2	100.0	1.422151	0.927972	-1.100715	0.779810	1.232191	2.093703	3.535038
att3	100.0	2.039149	1.046880	-1.321589	1.525548	2.152552	2.707411	3.850263
att4	100.0	2.126455	1.098754	-1.076313	1.511639	2.234408	2.865394	4.199145
att5	100.0	1.551616	0.954734	-1.247922	0.966292	1.746861	2.163355	3.720899
att6	100.0	0.762251	0.963111	-1.482391	0.022470	0.574297	1.502335	3.026452
att7	100.0	0.282647	1.096712	-1.598712	-0.603378	0.120230	1.226563	2.454195
att8	100.0	0.333630	0.877640	-1.508060	-0.464767	0.366122	1.008876	2.220327
att9	100.0	0.372973	0.761753	-1.609777	-0.075355	0.457255	0.990675	2.122435
att10	100.0	0.210769	0.739185	-1.600454	-0.205867	0.182365	0.776566	1.916524

2.2.3 Identifikasi Missing Value

print("=== Jumlah Missing Value ===")
missing_train = df_train.isnull().sum()
missing_test = df_test.isnull().sum()

print("Data Latih:")
if missing_train.sum() == 0:
    print("Tidak ada missing value.")
else:
    print(missing_train[missing_train > 0])

print("\nData Uji:")
if missing_test.sum() == 0:
    print("Tidak ada missing value.")
else:
    print(missing_test[missing_test > 0])

fig, ax = plt.subplots(1, 2, figsize=(12, 4))

# Heatmap hanya menunjukkan keberadaan missing (True = putih/merah, False = putih)
sns.heatmap(df_train.isnull(), cbar=False, cmap='Reds', ax=ax[0], yticklabels=False, xticklabels=False)
ax[0].set_title('Missing Value - Data Latih', fontsize=12)
ax[0].set_facecolor('white')

sns.heatmap(df_test.isnull(), cbar=False, cmap='Reds', ax=ax[1], yticklabels=False, xticklabels=False)
ax[1].set_title('Missing Value - Data Uji', fontsize=12)
ax[1].set_facecolor('white')

fig.patch.set_facecolor('white')
for a in ax:
    a.figure.set_facecolor('white')

plt.tight_layout()
plt.show()

=== Jumlah Missing Value ===
Data Latih:
Tidak ada missing value.

Data Uji:
Tidak ada missing value.

2.2.4 Distribusi Kelas

print("=== Distribusi Kelas ===")
print("Data Latih:")
print(df_train['target'].value_counts().sort_index())

print("\nData Uji:")
print(df_test['target'].value_counts().sort_index())


fig, ax = plt.subplots(1, 2, figsize=(10, 4))

sns.countplot(data=df_train, x='target', ax=ax[0], palette='pastel')
ax[0].set_title('Distribusi Kelas - Data Latih')

sns.countplot(data=df_test, x='target', ax=ax[1], palette='pastel')
ax[1].set_title('Distribusi Kelas - Data Uji')

plt.tight_layout()
plt.show()

=== Distribusi Kelas ===
Data Latih:
target
-1    31
 1    69
Name: count, dtype: int64

Data Uji:
target
-1    36
 1    64
Name: count, dtype: int64

/tmp/ipykernel_17518/1495402320.py:12: FutureWarning: 

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.countplot(data=df_train, x='target', ax=ax[0], palette='pastel')
/tmp/ipykernel_17518/1495402320.py:15: FutureWarning: 

Passing `palette` without assigning `hue` is deprecated and will be removed in v0.14.0. Assign the `x` variable to `hue` and set `legend=False` for the same effect.

  sns.countplot(data=df_test, x='target', ax=ax[1], palette='pastel')

2.2.5 Visualisasi Time Series Contoh Per Kelas

print("Jumlah sampel normal:", len(df_train[df_train['target'] == 1]))
print("Jumlah sampel abnormal:", len(df_train[df_train['target'] == -1]))

if len(df_train[df_train['target'] == 1]) == 0 or len(df_train[df_train['target'] == -1]) == 0:
    print("⚠️ Peringatan: Tidak ada data ditemukan untuk salah satu kelas!")
else:
    fig, axes = plt.subplots(2, 2, figsize=(12, 8))
    axes = axes.flatten()

    sample_normal = df_train[df_train['target'] == 1].iloc[:2]
    sample_abnormal = df_train[df_train['target'] == -1].iloc[:2]

    for i, row in enumerate(sample_normal.iterrows()):
        idx, data = row
        axes[i].plot(data[:-1], color='green', linewidth=2)
        axes[i].set_title(f'Normal Sample {i+1}')
        axes[i].grid(True, alpha=0.3)

    for i, row in enumerate(sample_abnormal.iterrows()):
        idx, data = row
        axes[i+2].plot(data[:-1], color='red', linewidth=2)
        axes[i+2].set_title(f'Abnormal Sample {i+1}')
        axes[i+2].grid(True, alpha=0.3)

    plt.tight_layout()
    plt.show()

Jumlah sampel normal: 69
Jumlah sampel abnormal: 31

2.2.6 Deteksi Outlier Menggunakan ABOD (Angle-Based Outlier Detection)

from pyod.models.abod import ABOD

df_combined = pd.concat([df_train, df_test], ignore_index=True)

X = df_combined.iloc[:, :-1].values 
y = df_combined['target'].values

abod = ABOD(contamination=0.05)  
outlier_labels = abod.fit_predict(X)

df_combined['outlier'] = outlier_labels

n_outliers = (outlier_labels == -1).sum()
print(f"🔍 Jumlah outlier terdeteksi: {n_outliers} dari {len(df_combined)} sampel")

fig, ax = plt.subplots(figsize=(10, 6))

normal_sample = df_combined[df_combined['outlier'] == 1].iloc[0, :-2].values
ax.plot(normal_sample, color='green', label='Normal Sample', linewidth=2)

if n_outliers > 0:
    outlier_sample = df_combined[df_combined['outlier'] == -1].iloc[0, :-2].values
    ax.plot(outlier_sample, color='red', label='Detected Outlier', linewidth=2, linestyle='--')
else:
    print("⚠️ Tidak ada outlier terdeteksi — hanya menampilkan sampel normal.")

ax.set_title('Contoh Time Series: Normal vs. Outlier (ABOD)')
ax.legend()
ax.grid(True, alpha=0.3)
plt.show()

/usr/local/python/3.12.1/lib/python3.12/site-packages/sklearn/utils/deprecation.py:87: FutureWarning: Function fit_predict is deprecated
  warnings.warn(msg, category=FutureWarning)

🔍 Jumlah outlier terdeteksi: 0 dari 200 sampel
⚠️ Tidak ada outlier terdeteksi — hanya menampilkan sampel normal.