private static void DrawPlot(IList <DayInfo> days, IList <DayInfo> anomalies)
{
days = days.Where(x => x.Date >= new DateTime(2017, 9, 1) && x.Date <= new DateTime(2017, 9, 30)).ToList();
anomalies = anomalies.Where(x => x.Date >= new DateTime(2017, 9, 1) && x.Date <= new DateTime(2017, 9, 30)).ToList();
using (var plot = new PLStream())
{
plot.sdev("pngcairo"); // png rendering
plot.sfnam("data.png"); // output filename
plot.spal0("cmap0_alternate.pal"); // alternate color palette
plot.init();
plot.env(
1, // x-axis range
days.Count,
0, // y-axis range
150,
AxesScale.Independent, // scale x and y independently
AxisBox.BoxTicksLabelsAxes); // draw box, ticks, and num ticks
plot.lab(
"Date", // x-axis label
"Count", // y-axis label
"Press releases September 2017"); // plot title
plot.line(
(from x in Enumerable.Range(1, days.Count) select(double) x).ToArray(),
(from p in days select(double) p.Count).ToArray());
// plot the spikes
plot.col0(2); // blue color
plot.schr(3, 3); // scale characters
plot.string2(
(from s in anomalies select(double) days.ToList().FindIndex(x => x.Date == s.Date) + 1).ToArray(),
(from s in anomalies select(double) s.Count + 15).ToArray(),
"↓");
plot.eop();
}
}
/// <summary>
/// The main program entry point.
/// </summary>
/// <param name="args">The command line parameters.</param>
static void Main()
{
// create the machine learning context
var context = new MLContext();
// load the data file
Console.WriteLine("Loading data...");
var dataView = context.Data.LoadFromTextFile <SalesRecord>(path: dataPath, hasHeader: true, separatorChar: ',');
// get an array of data points
var sales = context.Data.CreateEnumerable <SalesRecord>(dataView, reuseRowObject: false).ToArray();
// plot the data
var pl = new PLStream();
pl.sdev("pngcairo"); // png rendering
pl.sfnam("data.png"); // output filename
pl.spal0("cmap0_alternate.pal"); // alternate color palette
pl.init();
pl.env(
0, 36, // x-axis range
0, 800, // y-axis range
AxesScale.Independent, // scale x and y independently
AxisBox.BoxTicksLabelsAxes); // draw box, ticks, and num ticks
pl.lab(
"Date", // x-axis label
"Sales", // y-axis label
"Shampoo sales over time"); // plot title
pl.line(
(from x in Enumerable.Range(0, sales.Count()) select(double) x).ToArray(),
(from p in sales select(double) p.Sales).ToArray()
);
// build a training pipeline for detecting spikes
var pipeline = context.Transforms.DetectIidSpike(
outputColumnName: nameof(SalesPrediction.Prediction),
inputColumnName: nameof(SalesRecord.Sales),
confidence: 95,
pvalueHistoryLength: sales.Count() / 4); // 25% of x-range
// train the model
Console.WriteLine("Detecting spikes...");
var model = pipeline.Fit(dataView);
// predict spikes in the data
var transformed = model.Transform(dataView);
var predictions = context.Data.CreateEnumerable <SalesPrediction>(transformed, reuseRowObject: false).ToArray();
// find the spikes in the data
var spikes = (from i in Enumerable.Range(0, predictions.Count())
where predictions[i].Prediction[0] == 1
select(Day: i, Sales: sales[i].Sales));
// plot the spikes
pl.col0(2); // blue color
pl.schr(3, 3); // scale characters
pl.string2(
(from s in spikes select(double) s.Day).ToArray(),
(from s in spikes select(double) s.Sales + 40).ToArray(),
"!");
// build a training pipeline for detecting change points
var pipeline2 = context.Transforms.DetectIidChangePoint(
outputColumnName: nameof(SalesPrediction.Prediction),
inputColumnName: nameof(SalesRecord.Sales),
confidence: 95,
changeHistoryLength: sales.Count() / 4); // 25% of x-range
// train the model
Console.WriteLine("Detecting change points...");
var model2 = pipeline2.Fit(dataView);
// get predictions
transformed = model2.Transform(dataView);
predictions = context.Data.CreateEnumerable <SalesPrediction>(transformed, reuseRowObject: false).ToArray();
// find the change points in the data
var changes = (from i in Enumerable.Range(0, predictions.Count())
where predictions[i].Prediction[0] == 1
select(Day: i, Sales: sales[i].Sales));
// plot the change points as vertical red lines
pl.col0(3);
foreach (var c in changes)
{
pl.line(new double[] { c.Day, c.Day }, new double[] { 0, 800 });
}
pl.eop();
Console.WriteLine("Saved output file: data.png");
}
/// <summary>
/// The main program entry point.
/// </summary>
/// <param name="args">The command line parameters.</param>
static void Main()
{
// create the machine learning context
var context = new MLContext();
// load the data file
Console.WriteLine("Loading data...");
var dataView = context.Data.LoadFromTextFile <MeterData>(path: dataPath, hasHeader: true, separatorChar: ',');
// get an array of data points
var values = context.Data.CreateEnumerable <MeterData>(dataView, reuseRowObject: false).ToArray();
// plot the data
var pl = new PLStream();
pl.sdev("pngcairo"); // png rendering
pl.sfnam("data.png"); // output filename
pl.spal0("cmap0_alternate.pal"); // alternate color palette
pl.init();
pl.env(
0, 90, // x-axis range
0, 5000, // y-axis range
AxesScale.Independent, // scale x and y independently
AxisBox.BoxTicksLabelsAxes); // draw box, ticks, and num ticks
pl.lab(
"Day", // x-axis label
"Power consumption", // y-axis label
"Power consumption over time"); // plot title
pl.line(
(from x in Enumerable.Range(0, values.Count()) select(double) x).ToArray(),
(from p in values select(double) p.Consumption).ToArray()
);
// build a training pipeline for detecting spikes
var pipeline = context.Transforms.SsaSpikeEstimator(
nameof(SpikePrediction.Prediction),
nameof(MeterData.Consumption),
confidence: 98,
pvalueHistoryLength: 30,
trainingWindowSize: 90,
seasonalityWindowSize: 30);
// train the model
Console.WriteLine("Detecting spikes...");
var model = pipeline.Fit(dataView);
// predict spikes in the data
var transformed = model.Transform(dataView);
var predictions = context.Data.CreateEnumerable <SpikePrediction>(transformed, reuseRowObject: false).ToArray();
// find the spikes in the data
var spikes = (from i in Enumerable.Range(0, predictions.Count())
where predictions[i].Prediction[0] == 1
select(Day: i, Consumption: values[i].Consumption));
// plot the spikes
pl.col0(2); // blue color
pl.schr(3, 3); // scale characters
pl.string2(
(from s in spikes select(double) s.Day).ToArray(),
(from s in spikes select(double) s.Consumption + 200).ToArray(),
"↓");
pl.eop();
}