oguiza
commented
3 years ago Hi @jeznag ,
No need to apologize for being a noobie. We all are in one area or another :)
There are a few issues with your code. I've made some changes to make it work:
from datetime import datetime
epoch = datetime.utcfromtimestamp(0)
def unix_time_seconds(dt):
return (dt - epoch).total_seconds()
df_json = {"unix_timestamp":{"6":1451606400.0,"1558":1454284800.0,"3110":1456790400.0,"4662":1459468800.0,"6214":1462060800.0,"7766":1464739200.0,"9318":1467331200.0,"10870":1470009600.0,"12422":1472688000.0,"13974":1475280000.0,"15526":1477958400.0,"17078":1480550400.0,"18630":1483228800.0,"20182":1485907200.0,"21734":1488326400.0,"23286":1491004800.0,"24838":1493596800.0,"26390":1496275200.0,"27942":1498867200.0,"29494":1501545600.0,"31046":1504224000.0,"32598":1506816000.0,"34150":1509494400.0,"35702":1512086400.0,"37254":1514764800.0,"38806":1517443200.0,"40358":1519862400.0,"41910":1522540800.0,"43462":1525132800.0,"45014":1527811200.0,"46566":1530403200.0,"48118":1533081600.0,"49670":1535760000.0,"51222":1538352000.0,"52774":1541030400.0,"54326":1543622400.0,"55878":1546300800.0,"57430":1548979200.0,"58982":1551398400.0,"60534":1554076800.0,"62086":1556668800.0,"63638":1559347200.0,"65190":1561939200.0,"66742":1564617600.0,"68294":1567296000.0,"69846":1569888000.0,"71398":1572566400.0,"72950":1575158400.0,"74502":1577836800.0,"76054":1580515200.0,"77606":1583020800.0},"Usage_MWh":{"6":5.34858,"1558":3.78055,"3110":3.4831,"4662":3.74901,"6214":3.02347,"7766":7.63334,"9318":5.62975,"10870":5.51058,"12422":4.36067,"13974":3.29915,"15526":2.76066,"17078":2.94552,"18630":2.7777,"20182":2.76716,"21734":5.78573,"23286":4.8537129444,"24838":3.1271232778,"26390":2.8168842646,"27942":2.8774968882,"29494":2.8774968882,"31046":2.7846744079,"32598":2.8774968882,"34150":2.7846744079,"35702":2.8774968882,"37254":2.8774968882,"38806":2.5990294474,"40358":2.8774968882,"41910":2.5288257571,"43462":2.5724465738,"45014":3.3510199005,"46566":3.060722109,"48118":2.6989527056,"49670":2.5984900474,"51222":3.4421489889,"52774":3.4093083871,"54326":3.5249084516,"55878":3.1468401143,"57430":3.0175142015,"58982":3.3731491579,"60534":3.0313829708,"62086":3.1152347778,"63638":3.2681218106,"65190":3.4173398852,"66742":2.897951582,"68294":3.3056545,"69846":3.1457436,"71398":2.646408469,"72950":2.5245141129,"74502":2.7281552182,"76054":6.5999127071,"77606":7.7869288929}}
df = pd.DataFrame.from_dict(df_json)
window_length = 5
stride = 1
horizon = 1
X, y = SlidingWindow(window_length=window_length, stride=stride, horizon=horizon, get_x=['unix_timestamp'], get_y='Usage_MWh')(df)
splits = TimeSplitter()(X)
tfms = [None, [ToFloat(), ToNumpyTensor()]]
dsets = TSDatasets(X, y, tfms=tfms, splits=splits)
dls = TSDataLoaders.from_dsets(dsets.train, dsets.valid)
model = InceptionTime(dls.vars, 1)
learn = Learner(dls, model, loss_func=MSELossFlat())
learn.fit_one_cycle(100, 1e-3)
learn.plot_metrics()
valid_preds, valid_targets = learn.get_preds(ds_idx=1)
plt.plot(df.index[-len(splits[1]):].values, valid_preds, label='preds')
plt.plot(df.index[-len(splits[1]):].values, valid_targets, label='targets')
plt.title('Valid Preds vs Targets')
plt.legend(loc='best')
plt.show()With this you should get an output like this:
(tensor([3.3068, 3.4284, 3.5788, 3.7468, 3.9250, 4.1177, 4.3224, 4.5454, 4.7916]),
tensor([3.4173, 2.8980, 3.3057, 3.1457, 2.6464, 2.5245, 2.7282, 6.5999, 7.7869]))
The main issues you were facing are:
- You have trained a model to take the last 5 steps and predict the next. But then you were only passing one value to predict the next. It doesn't work this way. The input shape in the future needs to be the same as the input during training.
- The input was a numpy array, but then you were passing a tensor (hence your error message). You need to pass numpy arrays that you want to have predicted.
- It's easier to have a smaller mse error if you have lots of data. Since you have very few datapoints here, it'd be better to set stride to 1 so that your training set is a bit larger. Otherwise you have too few training samples.
Please, let me know if this resolves your issues.
jeznag
Sorry this is a really noob question. I've gone through tutorial 4 on time series regression and have successfully trained a learner. What I can't figure out is making a prediction on a future date. Here's what I tried:
I get
Sorry again for the really dumb question. It's probably a lack of fastai knowledge. I'll do a pull request afterwards and add the answer to the tutorial notebook.