tech.nmfin.algoquant.demo.model.infantino2010

Class Infantino2010ParamAnalysisDemo

java.lang.Object

tech.nmfin.algoquant.demo.model.infantino2010.Infantino2010ParamAnalysisDemo

@Demo
public class Infantino2010ParamAnalysisDemo
extends Object

Using this example. Uses SensitivityAnalyzer to evaluate Infantino strategy parameter sets on Yahoo history. The stock universe is filtered for data availability before simulation. Model fitting windows, forecast horizons and retained components have separate roles; parameter-search results are in-sample comparisons unless you add a holdout period.

Model and publication. The thesis combines PCA dimension reduction with regressions of future accumulated log returns on past accumulated principal-component returns; a separate regime decision can change how deviations are traded. This is a daily-bar adaptation, not a reproduction of its one-second quote experiment.

Implementation in this tutorial. Here T=500 is the rolling log-return sample length, k=2 is the retained dimension, N is the filtered IT-stock count, and H=1,21,...,201 is an accumulation/forecast horizon in observations. Regime switching is disabled. SensitivityAnalyzer creates an Infantino2010Strategy per parameter set; synchronized order books supply aligned prices. Infantino2010LongShortConstantValue reads signal.dReturns(), uses fixed per-stock notional and does not automatically invest simulator cash. The configured Commission measure records a statistic; it does not by itself install a broker fee scheme. Search results on the same interval are not holdout validation.

See Also:

L. R. Infantino and S. Itzhaki (2010), Developing high-frequency equities trading models, MIT MBA thesis

Constructor Summary

Constructors

Constructor and Description
Infantino2010ParamAnalysisDemo(List<Stock> stocks, dev.nm.misc.datastructure.time.LocalDateTimeInterval interval, ExchangeRateTable rates) Loads the chosen products and builds a simulator for the horizon comparisons.

Method Summary

Modifier and Type	Method and Description
static void	main(String[] args)
static void	printReport(SensitivityAnalysisReport<Infantino2010Strategy.Param> report) Prints every configuration and its account-level performance report.
SensitivityAnalysisReport<Infantino2010Strategy.Param>	run(List<Infantino2010Strategy.Param> params) Runs and plots each supplied model configuration on this instance's market data.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

Infantino2010ParamAnalysisDemo

public Infantino2010ParamAnalysisDemo(List<Stock> stocks,
                                      dev.nm.misc.datastructure.time.LocalDateTimeInterval interval,
                                      ExchangeRateTable rates)

Loads the chosen products and builds a simulator for the horizon comparisons.

Parameters:

stocks - products in a stable order shared by PCA columns and order sizing

interval - daily-history interval, including data needed for model warm-up

rates - currency-conversion table used to express performance consistently

Method Detail

main

public static void main(String[] args)
                 throws MapOfExceptions

Throws:

MapOfExceptions

run

public SensitivityAnalysisReport<Infantino2010Strategy.Param> run(List<Infantino2010Strategy.Param> params)
                                                           throws MapOfExceptions

Runs and plots each supplied model configuration on this instance's market data. A fresh strategy is constructed by the analyzer's factory for each configuration.

Parameters:

params - configurations to compare; changing H changes the forecast horizon

Returns:

report mapping each configuration to its simulated performance

Throws:

MapOfExceptions - if one or more parameter simulations fail

printReport

public static void printReport(SensitivityAnalysisReport<Infantino2010Strategy.Param> report)

Prints every configuration and its account-level performance report.

Parameters:

report - completed sensitivity results; values are not automatically ranked