{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Predict Temporal Series"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"collapsed": false
},
"outputs": [
{
"data": {
"text/html": [
""
],
"text/plain": [
""
]
},
"execution_count": 1,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"import addutils.toc ; addutils.toc.js(ipy_notebook=True)"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"collapsed": false
},
"outputs": [
{
"data": {
"text/html": [
"\n",
"\n"
],
"text/plain": [
""
]
},
"execution_count": 2,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"import scipy.io\n",
"import numpy as np\n",
"import pandas as pd\n",
"from addutils import css_notebook\n",
"css_notebook()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 1 Predict Univariate Temporal Data"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"This is the simpler case in temporal data prediction: we have just one measure to predict. The easier task is to made a one-step prediction. Here we test different ensamble methods on a multi-step prediction.\n",
"\n",
"Here for simplicity we consider data sampled with a constant sample time. in this way, when we speak about time, it can be either time or the number of the sample.\n",
"At every time-step the system will have an output **y(t)** and our goal is to predict the future output nk steps forward **y(t+nk)**.\n",
"\n",
"To fit the regressor we built a vector of **nk+1** past data samples: **[y(t-nk), ... , y(t-2), y(t-1), y(t)]**. Past measured values that build the regressor vector are referred as **Standard Regressor Components**. Is instead we define the values in the form of arbitrary user-defined functions of delayed input variables we call it **Custom Regressor Components**. In this case we will use just Standard Regressor Components.\n",
"\n",
"Usually predicting something many timesteps in advance is very hard: here we want to test the predictive capabilitied of many different ensemble regressors by predicting 9 timesteps in the future.\n",
"\n",
"Hint: try to change nb (the regressor vector length) to see how this affect the prediction"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"collapsed": false
},
"outputs": [],
"source": [
"%matplotlib inline\n",
"import matplotlib.pyplot as plt"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"collapsed": false
},
"outputs": [
{
"data": {
"text/html": [
" \n",
"\n",
"\n",
" \n",
"\n",
" \n",
"
\n",
"
BokehJS successfully loaded.\n",
"