private List <ChartTransaction> ComputeForecast(int id, int forecastMonths = 12)
{
List <ChartTransaction> transactions = new List <ChartTransaction>();
Brand currentBrand = Hardware.FirstOrDefault(x => x.ConfigId == id).Brand;
string currentModel = Hardware.FirstOrDefault(x => x.ConfigId == id).Model;
TimeSeriesPrediction prediction = new TimeSeriesPrediction(Transactions, currentBrand, currentModel);
prediction.GenerateFutureForecast(forecastMonths);
DateTime today = DateTime.Today;
foreach (Phone p in prediction.PhoneCollection.Phones)
{
DateTime currentDate = p.Date;
var datetime = DateTimeSpan.CompareDates(currentDate, today);
int yearDifference = datetime.Years;
if (yearDifference <= 2)
{
this.AddTransactionToRecord(id, new ChartTransaction {
Date = p.Date, Price = p.Forecast.Value
});
transactions.Add(new ChartTransaction
{
Date = p.Date,
Price = p.Forecast.Value
});
}
}
return(transactions);
}
private List <int> AutogenerateMissingDataIfPossible(List <int> selectedPhoneIds, int months = 3)
{
List <int> eligibleIds = selectedPhoneIds;
DateTime today = DateTime.Today;
int currentMonth = DateTime.Today.Month;
int currentYear = DateTime.Today.Year;
foreach (int id in selectedPhoneIds)
{
Brand selectedBrand = Hardware.FirstOrDefault(x => x.ConfigId == id).Brand;
string selectedModel = Hardware.FirstOrDefault(x => x.ConfigId == id).Model;
DateTime latestDate = Phones.Where(x => x.Brand == selectedBrand && x.Model == selectedModel).Max(x => x.Date);
var datetime = DateTimeSpan.CompareDates(latestDate, today);
int monthDifference = datetime.Years * 12 + datetime.Months;
// Remove id if latest date cannot be found or the gap is too big
if (latestDate == null || monthDifference > 3)
{
eligibleIds.RemoveAll(x => x == id);
Errors.Add($"{selectedBrand} {selectedModel} - Data for the last {monthDifference} months is unavailable.");
Errors = Errors.Distinct().ToList();
}
else
{
System.Diagnostics.Debug.WriteLine($"Generating data");
TimeSeriesPrediction prediction = new TimeSeriesPrediction(Phones, selectedBrand, selectedModel);
prediction.GenerateFutureForecast(monthDifference);
}
}
return(new List <int>());
}
static void Main(string[] args)
{
Console.WriteLine("**** Welcome to world of data science and machine learning using Microsoft .Net!!! ****");
#region EDA
Console.WriteLine("Performing Exploratory Data Analysis\n");
var exploratoryDataAnalysis = new ExploratoryDataAnalysis();
exploratoryDataAnalysis.Analyze();
Console.WriteLine("Exploratory Data Completed\n");
#endregion
#region Time Series Prediction
Console.WriteLine("\nPress any key to start Time Series Prediction\n");
Console.ReadKey();
var timeSeriesPrediction = new TimeSeriesPrediction();
timeSeriesPrediction.Forecast();
Console.WriteLine("Time Series Prediction Completed\n");
#endregion
Console.WriteLine("******* END *********");
Console.ReadLine();
}
// GET: Items
public async Task <IActionResult> Index()
{
List <Item> items = _phoneContext.Items.ToList();
TimeSeriesPrediction forecast = new TimeSeriesPrediction(items, Timeframe.Monthly);
forecast.GenerateFutureForecast(12);
return(View(await _phoneContext.Items.ToListAsync()));
}
static void Main(string[] args)
{
try
{
//Setting global device
Logging.OnWriteLog += Logging_OnWriteLog;
//XOR Example
XORExample.LoadData();
XORExample.BuildModel();
XORExample.Train();
//Housing regression example
HousingRegression.LoadData();
HousingRegression.BuildModel();
HousingRegression.Train();
//MNIST Classification example
MNISTClassifier.LoadData();
MNISTClassifier.BuildModel();
MNISTClassifier.Train();
//LSTM Time series example
TimeSeriesPrediction.LoadData();
TimeSeriesPrediction.BuildModel();
TimeSeriesPrediction.Train();
//Multi variate time series prediction
MiltiVariateTimeSeriesPrediction.LoadData();
MiltiVariateTimeSeriesPrediction.BuildModel();
MiltiVariateTimeSeriesPrediction.Train();
//Cifar - 10 Classification example
//Cifar10Classification.LoadData();
//Cifar10Classification.BuildModel();
//Cifar10Classification.Train();
//Image classification example
Console.WriteLine("ResNet50 Prediction: " + ImageClassification.ImagenetTest(SiaNet.Common.ImageNetModel.ResNet50)[0].Name);
//Console.WriteLine("Cifar 10 Prediction: " + ImageClassification.Cifar10Test(SiaNet.Common.Cifar10Model.ResNet110)[0].Name);
//Object Detection
ObjectDetection.PascalDetection();
//ObjectDetection.GroceryDetection();
Console.ReadLine();
}
catch (Exception ex)
{
Console.WriteLine(ex.ToString());
Console.ReadLine();
}
}
private List <ChartTransaction> ComputeForecast(int id, int forecastMonths = 12)
{
var asd = ContainsEnoughData(SelectedItems);
// A ChartTransaction object contains date of purchase and price.
List <ChartTransaction> transactions = new List <ChartTransaction>();
// Find brand and model based on passed id
Brand selectedBrand = Hardware.FirstOrDefault(m => m.ConfigId == id).Brand;
string selectedModel = Hardware.FirstOrDefault(m => m.ConfigId == id).Model;
// Generates time series forecast.
// RemoveUnforecastableIds method already checks if there are enough transactions to compute.
TimeSeriesPrediction prediction = new TimeSeriesPrediction(Phones, selectedBrand, selectedModel);
prediction.GenerateFutureForecast(forecastMonths);
DateTime today = DateTime.Today;
// Itterate through all objects
foreach (Phone p in prediction.PhoneCollection.Phones)
{
DateTime currentDate = p.Date;
var datetime = DateTimeSpan.CompareDates(currentDate, today);
int yearDifference = datetime.Years;
// Forecast is generated using ALL data
// Chart only displays data from the last 2 years + the forecast
if (yearDifference <= 2)
{
// Adds to the List<Dict<int, ChartTrans>> so calculations can be done
// In order to find the best/worst future price
this.AddTransactionToRecord(id, new ChartTransaction {
Date = p.Date, Price = p.Forecast.Value
});
// Records their date and price.
transactions.Add(new ChartTransaction
{
Date = p.Date,
Price = p.Forecast.Value
});
}
}
return(transactions);
}
