应用机器学习模型

用 Python 分析 IoT 数据

Matthias Voppichler

IT Developer

模型回顾

# 创建管道
pl = Pipeline([
        ("scale", StandardScaler()),
        ("logreg", LogisticRegression())
    ])

# 拟合管道
pl.fit(X_train, y_train)
print(pl.score(X_test, y_test))

0.8897932222860425

预测

predictions = pl.predict(X_test)

print(predictions)
print(f"测试集长度: {len(X_test)}")
print(f"预测长度: {len(predictions)}")

[0 0 0 ... 1 1 1]
测试集长度: 500
预测长度: 500

记录转换

print(single_record)

{'timestamp': '2018-11-30 18:15:00',
  'humidity': 81.7,
  'pressure': 1019.8,
  'temperature': 1.5},

cols = X_train.columns
df = pd.DataFrame.from_records([single_record], 
                               index="timestamp", 
                               columns=cols)

应用于数据流

def on_message(client, userdata, message):
    data = json.loads(message.payload)

    df = pd.DataFrame.from_records([data], 
                                   index="timestamp", 
                                   columns=cols)

    category = pl.predict(df)
    maybe_alert(category[0])

subscribe.callback(on_message, topic, hostname=MQTT_HOST)

让我们来练习！

用 Python 分析 IoT 数据