Demo machine learning cho việc dự đoán giá

Mô tả

SMP5
demo.py
Actions

Download data and upload model in google colab


import logging, urllib3, shutil
logging.basicConfig(level=logging.INFO)

def download_dataset(url, LOCAL_FILE_NAME): urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning) c = urllib3.PoolManager() with c.request("GET", url, preload_content=False) as res, open( LOCAL_FILE_NAME, "wb" ) as out_file: shutil.copyfileobj(res, out_file) logging.info("Dowload completed.")

logging.info("Started download script") URL = 'https://raw.githubusercontent.com/chanelcolgate/hydroelectric-project/19eb96f0d37b6dda92cceb21fde531ef9ff1f6c8/data/SMP_09-2021.csv' LOCAL_FILE_NAME = 'SMP_09-2021.csv' download_dataset(URL, LOCAL_FILE_NAME)

- Unzip model folder

!unzip "/content/saved_models-07-10.zip" -d "/"

- Predict
```python
import tensorflow as tf
from pandas import read_csv
from numpy import array
from math import sqrt
from sklearn.metrics import mean_squared_error
import numpy as np

# forecast with the fit model
def model_predict_2(model, history, n_input, n_features=1):
    # shape input for model
    x_input = array(history[-n_input:]).reshape((1, n_input, n_features))
    # make forecast
    yhat = model.predict(x_input, verbose=0)
    # correct forecast if it was differenced
    return yhat[0]

# root mean squared error or rmse
def measure_rmse(actual, predicted):
    return sqrt(mean_squared_error(actual, predicted))

model_path = '/content/saved_models/1633610713'
model = tf.keras.models.load_model(model_path)

series = read_csv(LOCAL_FILE_NAME, header=0, usecols=['SMP'])
data = series.values
# du doan 1 ngay, 0 ngay 1, 48 ngay 2
numberOfPredict = 1
periodOFPredict = numberOfPredict * 48
# dung ngay 5, 6 du doan ngay 7
indexOfPredict = 288 + periodOFPredict
actual = array(data[288:indexOfPredict]).reshape((periodOFPredict, 1)) # du lieu ngay 7 de so sanh
test = data[192:288] # du lieu ngay 5, 6 thang 9 nam 2021
test = array(test).reshape((1, 96))
predictions = list()
for i in range(0, periodOFPredict):
    prediction = model.predict(test, verbose=0)
    test = test.tolist()[0]
    test.extend(prediction.tolist()[0])
    test = array(test[-96:]).reshape((1, 96))
    predictions.extend(prediction.tolist()[0])
predictions = [1503.5 if num > 1503.5 else num for num in predictions] # gioi han tren
predictions = [0.0 if num < 0.0 else num for num in predictions] # gioi han duoi
predictions = [round(num, 2) for num in predictions]
measure_rmse(actual, predictions)

Add smp.py


%%writefile smp.py
import pandas as pd # type: ignore[import]
import numpy as np # type: ignore[import]
from tqdm import tqdm # type: ignore[import]
from pandas import read_excel
from numpy import array
from typing import List, Any, Tuple

class Smp: def init( self, fileName: str = "Bảng dự kiến chào giá T9.xlsx", sheetName: List[str] = ['07-09'], index: int = 0, A: List[str] = ['0', '5.5'], n: int = 2, priceContract : float = 917.220, expectedPrice: List[float] = [0.0]) -> None: df = read_excel( fileName, sheet_name = sheetName, skiprows = 3, nrows = 48, usecols = "A:J") self.data = df[sheetName[index]] self.A = A self.n = n self.priceContract = priceContract self.expectedPrice = expectedPrice

def strings(self) -> Any:
    index_of = {x: i for i, x in enumerate(self.A)}
    s = [self.A[0]] * self.n
    while True:
        yield ','.join(s)
        for i in range(1, self.n + 1):
            if s[-i] == self.A[-1]:
                s[-i] = self.A[0]
            else:
                s[-i] = self.A[index_of[s[-i]] + 1]
                break
        else:
            break

def revenue(
        self,
        outputContract: Any,
        priceContract: float,
        expectedPrice: Any,
        expectedOutput: Any,
        priceCan: Any) -> Tuple[float, float]:
    expectedRevenue = outputContract*(priceContract - expectedPrice - priceCan) + \
                      expectedOutput*(expectedPrice + priceCan)
    expectedRevenueContract = expectedOutput*priceContract
    if expectedRevenueContract.sum():
        result = expectedRevenue.sum()/expectedRevenueContract.sum() * 100 - 100
    else:
        result = -100.0
    return (expectedRevenueContract.sum(), result)

@property
def scores(self) -> List[Any]:
    configs = self.strings()
    return sorted((
            self.revenue(
                outputContract = self.data['Sản lượng hợp đồng (Qc)'].values,
                priceContract = self.priceContract,
                expectedPrice = array(self.expectedPrice),
                expectedOutput = array([float(e) for e in row.split(',') for _ in (0, 1)]),
                priceCan = self.data['Giá CAN'].values
            ),
            str([float(e) for e in row.split(',') for _ in (0, 1)])
        )
        for row in tqdm(configs, total=pow(2, self.n))
    )

- Import smp and predict to expected output. Ở đây đầu vào là ngày 7 tháng 9 và công suất là 11 MW
```python
from smp import Smp

smp = Smp(expectedPrice = predictions, sheetName = ['07-09'], A = ['0', '11'], n=24)
scores = smp.scores

Lấy 1000 kết quả có doanh thu cao nhất để sắp xếp theo tỉ lệ giữa doanh thu dự kiến theo thị trường và doanh thu dự kiến theo hợp đồng cao nhất
```
scores_1000 = scores[-1000:]
scores_1000.sort(key=lambda tup: tup[0][1])
scores_1000[-3:]
```
Chạy code demo.py để tạo ra file excel
```
uvicorn test:app
```
Lên localhost:8000/docs nhập các đầu vào

chanelcolgate / hydroelectric-project

Demo machine learning cho việc dự đoán giá #23

Mô tả

Actions