![\"time](\"img/tser.png\")
\n",
"\n",
"An example TSER problem is shown in the below image of soil spectrograms which can be used to estimate\n",
"the potassium concentration. Ground truth is found through expensive\n",
"lab based experiments that take some time. Spectrograms (ordered data series we treat as time series)\n",
"are cheap to obtain and the data can be collected in any environment. An accurate regressor from spectrogram\n",
"to concentration would make land and crop management more efficient.\n",
"\n",
"![\"spectrograph](\"img/spectra.png\")
"
],
"metadata": {
"collapsed": false
}
},
{
"cell_type": "markdown",
"source": [
"## Data Storage and Problem Types\n",
"\n",
"Regressors take time series input as either 3D numpy of shape`(n_cases, n_channels, n_timepoints)` for\n",
"equal length series or as a list of 2D numpy of length `[n_cases]`. All regressors work with equal\n",
"length problems. Regression functionality for unequal length is currently limited.\n",
"\n",
"`aeon` ships two example regression problems in the `datasets` module:"
],
"metadata": {
"collapsed": false
}
},
{
"cell_type": "code",
"execution_count": 9,
"outputs": [
{
"data": {
"text/plain": "(140, 1, 84)"
},
"execution_count": 9,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"import warnings\n",
"\n",
"import matplotlib.pyplot as plt\n",
"import numpy as np\n",
"\n",
"from aeon.datasets import load_cardano_sentiment, load_covid_3month\n",
"\n",
"warnings.filterwarnings(\"ignore\")\n",
"\n",
"covid_train, covid_train_y = load_covid_3month(split=\"train\")\n",
"covid_test, covid_test_y = load_covid_3month(split=\"test\")\n",
"cardano_train, cardano_train_y = load_cardano_sentiment(split=\"train\")\n",
"cardano_test, cardano_test_y = load_cardano_sentiment(split=\"test\")\n",
"covid_train.shape"
],
"metadata": {
"collapsed": false,
"ExecuteTime": {
"end_time": "2023-08-23T19:10:31.722451500Z",
"start_time": "2023-08-23T19:10:31.704499Z"
}
}
},
{
"cell_type": "markdown",
"source": [
"Covid3Month is from the [monash tser archive](http://tseregression.org/) who in turn got\n",
"the data from [WHO's COVID-19 database](https://covid19.who.int/). The goal of this dataset is to predict\n",
"COVID-19's death rate on 1st April 2020 for each country using daily confirmed cases for the last three months.\n",
"\n",
"This dataset contains 201 time series (140 train, 61 test), where each time series is\n",
"the daily confirmed cases for a country. The data is univariate, each series length 84."
],
"metadata": {
"collapsed": false
}
},
{
"cell_type": "code",
"execution_count": 10,
"outputs": [
{
"data": {
"text/plain": "[| \n | name | \nestimator | \n
|---|---|---|
| 0 | \nCNNRegressor | \n<class 'aeon.regression.deep_learning.cnn.CNNR... | \n
| 1 | \nCanonicalIntervalForestRegressor | \n<class 'aeon.regression.interval_based._cif.Ca... | \n
| 2 | \nCatch22Regressor | \n<class 'aeon.regression.feature_based._catch22... | \n
| 3 | \nDrCIFRegressor | \n<class 'aeon.regression.interval_based._drcif.... | \n
| 4 | \nDummyRegressor | \n<class 'aeon.regression._dummy.DummyRegressor'> | \n
| 5 | \nFCNRegressor | \n<class 'aeon.regression.deep_learning.fcn.FCNR... | \n
| 6 | \nFreshPRINCERegressor | \n<class 'aeon.regression.feature_based._fresh_p... | \n
| 7 | \nInceptionTimeRegressor | \n<class 'aeon.regression.deep_learning.inceptio... | \n
| 8 | \nIndividualInceptionRegressor | \n<class 'aeon.regression.deep_learning.inceptio... | \n
| 9 | \nIntervalForestRegressor | \n<class 'aeon.regression.interval_based._interv... | \n
| 10 | \nKNeighborsTimeSeriesRegressor | \n<class 'aeon.regression.distance_based._time_s... | \n
| 11 | \nRandomIntervalRegressor | \n<class 'aeon.regression.interval_based._interv... | \n
| 12 | \nRandomIntervalSpectralEnsembleRegressor | \n<class 'aeon.regression.interval_based._rise.R... | \n
| 13 | \nRegressorPipeline | \n<class 'aeon.regression.compose._pipeline.Regr... | \n
| 14 | \nResNetRegressor | \n<class 'aeon.regression.deep_learning.resnet.R... | \n
| 15 | \nRocketRegressor | \n<class 'aeon.regression.convolution_based._roc... | \n
| 16 | \nSklearnRegressorPipeline | \n<class 'aeon.regression.compose._pipeline.Skle... | \n
| 17 | \nTapNetRegressor | \n<class 'aeon.regression.deep_learning.tapnet.T... | \n
| 18 | \nTimeSeriesForestRegressor | \n<class 'aeon.regression.interval_based._tsf.Ti... | \n
| \n | name | \nestimator | \n
|---|---|---|
| 0 | \nCNNRegressor | \n<class 'aeon.regression.deep_learning.cnn.CNNR... | \n
| 1 | \nCanonicalIntervalForestRegressor | \n<class 'aeon.regression.interval_based._cif.Ca... | \n
| 2 | \nCatch22Regressor | \n<class 'aeon.regression.feature_based._catch22... | \n
| 3 | \nDrCIFRegressor | \n<class 'aeon.regression.interval_based._drcif.... | \n
| 4 | \nDummyRegressor | \n<class 'aeon.regression._dummy.DummyRegressor'> | \n
| 5 | \nFCNRegressor | \n<class 'aeon.regression.deep_learning.fcn.FCNR... | \n
| 6 | \nFreshPRINCERegressor | \n<class 'aeon.regression.feature_based._fresh_p... | \n
| 7 | \nInceptionTimeRegressor | \n<class 'aeon.regression.deep_learning.inceptio... | \n
| 8 | \nIndividualInceptionRegressor | \n<class 'aeon.regression.deep_learning.inceptio... | \n
| 9 | \nIntervalForestRegressor | \n<class 'aeon.regression.interval_based._interv... | \n
| 10 | \nKNeighborsTimeSeriesRegressor | \n<class 'aeon.regression.distance_based._time_s... | \n
| 11 | \nRandomIntervalRegressor | \n<class 'aeon.regression.interval_based._interv... | \n
| 12 | \nRandomIntervalSpectralEnsembleRegressor | \n<class 'aeon.regression.interval_based._rise.R... | \n
| 13 | \nResNetRegressor | \n<class 'aeon.regression.deep_learning.resnet.R... | \n
| 14 | \nRocketRegressor | \n<class 'aeon.regression.convolution_based._roc... | \n
| 15 | \nTapNetRegressor | \n<class 'aeon.regression.deep_learning.tapnet.T... | \n
| 16 | \nTimeSeriesForestRegressor | \n<class 'aeon.regression.interval_based._tsf.Ti... | \n