Pytorch predict Result is Almost the same

vision
1

I have only seven characteristics in a data forecasting the amount of bicycles. Why do I predict that the results are almost the same?
image
Out:
tensor([[50.9622],
[50.7544],
[50.2624],
[49.4988],
[50.6219],
[50.3502],
[50.2156],
[50.7703],
[51.0407],
[50.7978]])

2

Maybe your model is too simple. Better you show your full code.

3

Thank you

This is all the code

If you need all the data, I can give it to you.

import pandas as pd    
import numpy as np    
import matplotlib.pyplot as plt    
import torch˽    
import torch.nn as nn    
import warnings    
warnings.filterwarnings('ignore')    

train = pd.read_csv("../../Data/3-sofa公共自行车使用量预测/train.csv",index_col='id')    
test = pd.read_csv("../../Data/3-sofa公共自行车使用量预测/test.csv",index_col='id')    
submit = pd.read_csv("../../Data/3-sofa公共自行车使用量预测/sample_submit.csv")    

train.columns

OUt
Index([‘city’, ‘hour’, ‘is_workday’, ‘weather’, ‘temp_1’, ‘temp_2’, ‘wind’, ‘y’], dtype=‘object’)

train.head()

image

# 取出训练集的y
y_train = train.pop('y')

train = torch.FloatTensor(train.values)
test = torch.FloatTensor(test.values)
y_train = torch.FloatTensor(y_train.values)

net = torch.nn.Sequential(
    torch.nn.Linear(7,100),
    torch.nn.Sigmoid(),
    torch.nn.Linear(100,1),
)
optimizer = torch.optim.Adam(net.parameters(),lr=0.02)
loss_func = torch.nn.MSELoss()

for i in range(60):
    prediction = net(train)
    loss = loss_func(prediction,y_train)
    # 梯度归零
    optimizer.zero_grad()
    # 计算梯度
    loss.backward()
    # 更新结点
    optimizer.step()
    if i % 20 == 0:
        print(loss)

OUt˽˽

tensor(4836.1333, grad_fn=<MseLossBackward>)
tensor(2624.3867, grad_fn=<MseLossBackward>)
tensor(2307.5830, grad_fn=<MseLossBackward>)

y_train[:10]

tensor([15., 48., 21., 11., 39., 12., 11., 67., 77.,  2.])

net(train[:10])

Out
tensor([[51.4017],
[49.5164],
[48.2066],
[49.3453],
[49.4258],
[48.9165],
[48.8166],
[49.5939],
[50.7129],
[49.8309]])