Dotnet

Predicting BS using ML.Net Machine learning is typically the realm for R/Python. But can .Net move into this space? In this talk we will run through Microsoft’s new ML.Net framework including what it currently offers, how to build a learning pipeline and how to deploy a model to an Azure Service. https://slides.com/chrismckelt/deck-451cdb94-a37d-47b7-9d49-6686065e7d03

In this post we will build a prediction model using ML .Net to classify an individual bank transaction into a pre-defined group using the description text. We will then evaluate the effectiveness of the model using standard evaluation metrics. The steps involved to create a model using ML.Net are: Decide what type of machine learning problem we are trying to solve? Retrieve and clean the data (CSV/TSV used with ML.Net for no) Create a prediction service API Create a ‘Learning Pipeline’ to iterate over the model build process Training Testing Evaluating result metrics Exploring the output model file ...

So what are we trying to do here? Hypothesis Given a users bank statement we will be able to predict (within a statistical confidence level ) the transactions within a period into categories & sub categories for spending classification. Target Categories Banking & Finance Entertainment Food & Drinks Groceries Health & Beauty Holiday & Travel Home Household Utilities Income Insurance Kids Miscellaneous Shopping Transferring Money Transport Work & Study Sub categories available on this link ...

The proposed solution sequence flow follows: Posts in this series Charge Id – scratching the tech itch Charge Id – lean canvas Charge Id – solution overview Charge Id – analysing the data Charge Id – the prediction model Charge Id – deploying a ML.Net Model to Azure Code https://github.com/chrismckelt/vita

Lean Canvas Problem This is the start of a series of blog posts where I try to build an online personal budgeting system. The motivation for this was me trying to do a household budget usin With Open Banking approaching a common service to classify income and expenditure for bank statement transactions is not currently available on the market. The ability for a consumer to categorise their transactions sits within varied personal finance providers (eg. pocketbook, mint). ...

This is the start of a series of blog posts where I try to build an online personal budgeting system. The motivation for this was me trying to do a household budget using existing tools (Pocketbook, CashDesk, QuickBooks, Xero). Each time I needed to classify a large number of transactions manually (except Pocketbook they have got a pretty good classification engine) As I do lots of transfer between accounts the summaries never made sense. So I thought Id give it a go to build my own personal finance budgeter. ...

1. Add the buildProvider config for razor inside the compilation element 2. Application start –> Registers the html extension with razor 3. Create a Start.cshtml page and ingest the index.html page <compilation debug\="true" targetFramework\="4.6" \> <buildProviders\> <add extension\=".html" type\="System.Web.WebPages.Razor.RazorBuildProvider"/> </buildProviders\> </compilation\> System.Web.Razor.RazorCodeLanguage.Languages.Add("html", new CSharpRazorCodeLanguage()); WebPageHttpHandler.RegisterExtension("html"); @{ Layout \= null; @RenderPage("~/index.html") } <!-- Version + @System.Reflection.Assembly.GetAssembly(typeof (Startup)).GetName().Version.ToString(); \-->

[TestFixture] public class RouteTestFixture { [Test] public void Routes _should _not _change() { const string good = @ "all existing routes go here (hint run the test first then copy output to here)"; var goodRoutes = good.Split(Convert.ToChar(";")); var notFound = new List < string > (); var ass = Assembly.GetAssembly(typeof (EventStatisticsController)); var sb = new StringBuilder(); foreach(var type in ass.GetTypes()) { try { var members = type.GetMembers(); for (int i = 0; i < members.Length; i++) { if (members[i].IsDefined(typeof (RouteAttribute), false)) { Object[] atts = members[i].GetCustomAttributes(typeof (RouteAttribute), false); for (int j = 0; j < atts.Length; j++) { var routeAttribute = (RouteAttribute) atts[j]; string route = routeAttribute.Template + @ "/" + routeAttribute.Name; Console.WriteLine(route); sb.Append(route + ";"); if (!goodRoutes.Contains(route)) { notFound.Add(route); } } } } } catch (Exception e) { Console.WriteLine(@ "An exception occurred: {0}", e.Message); } } if (notFound.Any()) { Console.WriteLine(@ "-- MISSING ROUTES --"); foreach(var nf in notFound) { Console.WriteLine(nf); } Assert.Fail("Missing Routes: " + notFound.Count); } else { Assert.True(true, "All routes matched"); } Console.WriteLine(@ "------------------------------------------------"); Console.WriteLine(@ "Match the following string for future tests"); Console.WriteLine(sb.ToString()); } }

How to use a custom NHibernate User Type To store the state of a Risk (ie RiskState) as a value in a column we can use a NHibernate.UserType The RiskState can be one of the following: (these are all derived from RiskState which implements IRiskState) Our user type public class RiskStateNameUserType: IUserType { #region IUserType Members public object Assemble(object cached, object owner) { return cached; } public object DeepCopy(object value) { return value; } public object Disassemble(object value) { return value; } public int GetHashCode(object x) { return x.GetHashCode(); } public bool IsMutable { get { return false; } } public object NullSafeGet(System.Data.IDataReader dr, string\[\] names, object owner) { var property0 = NHibernateUtil.String.NullSafeGet(dr, names\[0\]); if (property0 == null) { return null; } IRiskState state; if (owner is Risk) { state = (IRiskState) Activator.CreateInstance(Type.GetType(typeof (IRiskState).Namespace + "." + (string) property0), owner); } else { state = (IRiskState) Activator.CreateInstance(Type.GetType(typeof (IRiskState).Namespace + "." + (string) property0)); } return state; } public void NullSafeSet(System.Data.IDbCommand cmd, object value, int index) { if (value == null) { ((IDataParameter) cmd.Parameters\[index\]).Value = DBNull.Value; } else { var state = (IRiskState) value; ((IDataParameter) cmd.Parameters\[index\]).Value = state.GetType().Name; } } public object Replace(object original, object target, object owner) { return original; } public Type ReturnedType { get { return typeof (RiskState); } } public NHibernate.SqlTypes.SqlType\[\] SqlTypes { get { return new\ [\] { NHibernateUtil.String.SqlType }; } } public new bool Equals(object x, object y) { if (x == null && y == null) return true; if (x == null || y == null) return false; return x.GetType() == y.GetType(); } #endregion } Using this on a property as follows: ...

A simple implemention for calling asych methods from the UI Examples dispatcher.ExecuteOnMainUIThread(CommandManager.InvalidateRequerySuggested); dispatcher.Execute(() => { SomeLongRunningMethodHere(); }); The interface public interface IDispatcher { void Execute(Action action); void ExecuteOnMainUIThread(Action action); } } Synchronous for use in Testing public class SynchronousDispatcher: IDispatcher { public void Execute(Action action) { action(); } public void ExecuteOnMainUIThread(Action action) { action(); } } Asynchronous for use by the application at run time public class AsynchronousDispatcher: IDispatcher { public void Execute(Action action) { action.BeginInvoke(CallBack, action); } public void ExecuteOnMainUIThread(Action action) { Dispatcher dispatcher; if (Application.Current != null) { dispatcher = Application.Current.Dispatcher; } else { dispatcher = Dispatcher.CurrentDispatcher; } dispatcher.Invoke(action); } private void CallBack(IAsyncResult result) { try { ((Action) result.AsyncState).EndInvoke(result); } catch (Exception ex) { // Need to raise the exception on the main thread ExecuteOnMainUIThread(() => { throw ex; } ); } finally { result.AsyncWaitHandle.Close(); } } }

using(var stream = Assembly.GetAssembly(typeof (StubPolicy)).GetManifestResourceStream( "Documents.TestHelpers.Files.test.msg")) { const int bufferLength = 256; var buffer = new Byte\[bufferLength\]; if (stream != null) { int bytesRead = stream.Read(buffer, 0, bufferLength); using(var fs = new FileStream(filename, FileMode.CreateNew, FileAccess.Write)) { // Write out the input stream while (bytesRead > 0) { fs.Write(buffer, 0, bytesRead); bytesRead = stream.Read(buffer, 0, bufferLength); } fs.Close(); } stream.Close(); } }

const int maxSize = 52428800; var binding = new BasicHttpBinding(); binding.Security.Mode = BasicHttpSecurityMode.TransportCredentialOnly; binding.Security.Transport.ClientCredentialType = HttpClientCredentialType.Windows; binding.Security.Transport.ProxyCredentialType = HttpProxyCredentialType.None; binding.Security.Message.ClientCredentialType = BasicHttpMessageCredentialType.UserName; binding.Security.Message.AlgorithmSuite = SecurityAlgorithmSuite.Default; binding.MaxReceivedMessageSize = 1000000; binding.CloseTimeout = new TimeSpan(0, 1, 0); binding.OpenTimeout = new TimeSpan(0,1,0); binding.ReceiveTimeout = new TimeSpan(0,10,0); binding.SendTimeout = new TimeSpan(0,1,0); binding.AllowCookies = false; binding.BypassProxyOnLocal = false; binding.HostNameComparisonMode = HostNameComparisonMode.StrongWildcard; binding.MaxBufferSize = maxSize; binding.MaxBufferPoolSize = maxSize; binding.MaxReceivedMessageSize = maxSize; binding.MessageEncoding = WSMessageEncoding.Mtom; binding.TextEncoding = Encoding.UTF8; binding.TransferMode = TransferMode.Buffered; binding.UseDefaultWebProxy = true; binding.ReaderQuotas.MaxDepth = 32; binding.ReaderQuotas.MaxStringContentLength = maxSize; binding.ReaderQuotas.MaxArrayLength = maxSize; binding.ReaderQuotas.MaxBytesPerRead = maxSize; binding.ReaderQuotas.MaxNameTableCharCount = maxSize; container = new IocContainer(LifestyleType.Transient); container.AddFacility<WcfFacility>().Register( Component .For<ISharePointFacadeService>() .Named("DmsGateway") .ActAs( new DefaultClientModel() { Endpoint = WcfEndpoint .BoundTo(binding) .At("http://localhost/SharepointFacade/DMSService.svc/mex") }));

public static class Extensions { public static Hashtable ConvertPropertiesAndValuesToHashtable(this object obj) { var ht = new Hashtable(); // get all public static properties of obj type var propertyInfos = obj.GetType().GetProperties().Where(a=>a.MemberType.Equals(MemberTypes.Property)).ToArray(); // sort properties by name Array.Sort(propertyInfos, (propertyInfo1, propertyInfo2) => propertyInfo1.Name.CompareTo(propertyInfo2.Name)); // write property names foreach (PropertyInfo propertyInfo in propertyInfos) { ht.Add(propertyInfo.Name, propertyInfo.GetValue(obj, BindingFlags.Public, null, null, CultureInfo.CurrentCulture)); } return ht; } } Tests using System; using System.Collections; using System.Globalization; using System.Linq; using System.Reflection; ...

In the web.config in between the tags <system.diagnostics> <sources> <source name=“System.ServiceModel” switchValue=“Information, ActivityTracing” propagateActivity=“true”> <listeners> <add name=“traceListener” type=“System.Diagnostics.XmlWriterTraceListener” initializeData= “WCFTrace.svclog” /> </listeners> </source> </sources> <trace autoflush=“true” /> </system.diagnostics>

using System; using System.Collections.Generic; using System.Linq; using System.Reflection; using System.Text; namespace TestHelpers { public static class RandomHelper { /* This method can be used to fill all public properties of an object with random values depending on their type. CAUTION: it does not fill attributes that end with 'ID' or attributes which are called 'pk'. They have to be filled manually.*/ private static readonly Random randomSeed = new Random(); /* Generates a random string with the given length*/ public static T FillPropertiesWithRandomValues<T>(bool fillBaseObjects) { return CreateItem<T>(true); } public static T FillPropertiesWithRandomValues<T>() { return CreateItem<T>(false); } private static T CreateItem<T>(bool fillBaseObjects) { Type type = typeof (T); var item = (T) Activator.CreateInstance(typeof (T)); while (type != null) { PropertyInfo[] infos = type.GetProperties(BindingFlags.Public | BindingFlags.Instance | BindingFlags.DeclaredOnly); if (infos.Length == 0) { if (item.GetType().BaseType != null) { infos = item.GetType().BaseType.GetProperties(BindingFlags.Public | BindingFlags.Instance | BindingFlags.DeclaredOnly); } } foreach (PropertyInfo info in infos) { Type infoType = info.PropertyType; Type nullableType = null; if (infoType.IsGenericType && infoType.GetGenericTypeDefinition().Equals(typeof (Nullable<>))) { nullableType = ExtractTypeFromNullable(infoType); } if (!info.CanWrite) continue; if (infoType.Equals(typeof (DateTime)) || infoType.Equals(typeof (DateTime?))) { info.SetValue(item, RandomDateTime(DateTime.Now, new DateTime(3000, 01, 01)), null); } else if (infoType.Equals(typeof (string))) { info.SetValue(item, info.Name + "_" + RandomString(20, false), null); } else if ((infoType.Equals(typeof (long)) || infoType.Equals(typeof (double)) || infoType.Equals(typeof (int)) || infoType.Equals(typeof (long)) || infoType.Equals(typeof (int))) && !info.Name.ToLower().EndsWith("id") && !info.Name.ToLower().Equals("pk")) { info.SetValue(item, RandomNumber(0, 999999), null); } else if ((infoType.Equals(typeof (long?)) || infoType.Equals(typeof (double?)) || infoType.Equals(typeof (int?)) || infoType.Equals(typeof (long?)) || infoType.Equals(typeof (int?))) && !info.Name.ToLower().EndsWith("id") && !info.Name.ToLower().Equals("pk")) { Type genericType = info.PropertyType.GetGenericArguments()[0]; info.SetValue(item, Convert.ChangeType(RandomNumber(0, 999999), genericType), null); } else if (infoType.Equals(typeof (decimal))) { info.SetValue(item, 1m, null); } else if (infoType.Equals(typeof (bool))) { info.SetValue(item, true, null); } else if (infoType.Equals(typeof (Guid))) { info.SetValue(item, Guid.NewGuid(), null); } else if (infoType.Equals(typeof (Enum))) { Array values = Enum.GetValues(infoType); List<object> list = values.Cast<object>().ToList(); object val = list[new Random().Next(0, list.Count)]; info.SetValue(item, val, null); } else if (((infoType.BaseType != null) && (infoType.BaseType.Equals(typeof (Enum))))) { Array values = Enum.GetValues(infoType); List<object> list = values.Cast<object>().ToList(); object val = list[new Random().Next(0, list.Count)]; info.SetValue(item, val, null); } else if (nullableType != null) { if (nullableType.BaseType.Equals(typeof (Enum))) { Array values = Enum.GetValues(nullableType); List<object> list = values.Cast<object>().ToList(); object val = list[new Random().Next(0, list.Count)]; info.SetValue(item, val, null); } } else if (infoType.IsArray) { Console.WriteLine("Object contains array of objects need to fill these"); MethodInfo sm = info.GetSetMethod(true); if (sm.ReturnType.IsArray) { object arrayObject = sm.Invoke(item, null); foreach (object element in (Array) arrayObject) { foreach (PropertyInfo arrayObjPinfo in element.GetType().GetProperties()) { Console.WriteLine(arrayObjPinfo.Name + ":" + arrayObjPinfo.GetGetMethod().Invoke(element, null)); } } } } } if (!fillBaseObjects) break; type = type.BaseType; } return item; } ///<summary>Indentify and extracting type from Nullable Type</summary> public static Type ExtractTypeFromNullable(Type type) { if (type.IsGenericType && type.GetGenericTypeDefinition() == typeof (Nullable<>)) { PropertyInfo valueProp = type.GetProperty("Value"); return valueProp.PropertyType; } else { return null; } } public static string RandomString(int size, bool lowerCase) { var randomString = new StringBuilder(size); /* Ascii start position (65 = A / 97 = a)*/ int start = lowerCase ? 97 : 65; /* Add random chars*/ for (int i = 0; i < size; i++) { randomString.Append((char) ((26*randomSeed.NextDouble()) + start)); } return randomString.ToString(); } public static int RandomNumber(int minimal, int maximal) { return randomSeed.Next(minimal, maximal); } /* Returns a random boolean value*/ public static bool RandomBool() { return randomSeed.NextDouble() > 0.5; } /* Returns a random color*/ public static DateTime RandomDateTime(DateTime min, DateTime max) { if (max <= min) { const string message = "Max must be greater than min."; throw new ArgumentException(message); } long minTicks = min.Ticks; long maxTicks = max.Ticks; double rn = ((Convert.ToDouble(maxTicks) - Convert.ToDouble(minTicks))*randomSeed.NextDouble()) + Convert.ToDouble(minTicks); return new DateTime(Convert.ToInt64(rn)); } } }