Want to create interactive content? It’s easy in Genially!

Get started free

Specifiche Prodotti Qi4M

Michelangelo Ferrazza

Created on February 23, 2024

Start designing with a free template

Discover more than 1500 professional designs like these:

Puzzle Game

Scratch and Win

Microlearning: How to Study Better

Branching Scenarios Challenge Mobile

Branching Scenario Mission: Innovating for the Future

Piñata Challenge

Teaching Challenge: Transform Your Classroom

Explore all templates

Transcript

User Journey

In an ever-changing financial landscape, the adoption of customized financial strategies and the integration of artificial intelligence (AI) models represent a real competitive advantage

This guide presents a flexible pipeline that allows clients to customize every aspect of their investment strategy, from investment universe selection to implementation and performance monitoring.

Investment Strategy

Performance Evaluation

Universe

Asset Class

Start

1. Asset Class

Equity
Each available asset class offers a unique set of differing options, which allows for diversification of the portfolio, risk control, and increasing potential profits.
Equity Index / ETFs
Fixed Income
Commodities
Crypto

2. Equity Universe

Sector
Basket Universe from Market Index
Geographical Area
Custom Thematic Basket via Qi4M Algorithm (1, 2, 3...)
Import Custom Universe
Market Filters
Metaverse
AI
Sport
Automotive
Hydrogen
Green Food
Travel
Customize
Fashion
Green Food
Travel
Eurostoxx
MSCI World
SP500
Regional
National
Global
Energy
Health Care
Financials
Industrials
Utilities
Technology
Technology
Materials
ESG score
Market Cap
Volume
P/E Ratio
Dividend Yield
Universe

3. Equity's Investment Strategy

WEIGHTING

Time Horizons

Equal/Cap Weight
Daily
Weekly
Risk Parity
Timing
Picking
Minimum Volatility
Monthly
Quarterly
Custom Optimization

3. Picking

Long
The investor selects specific securities or assets that are expected to increase in value over time.
Short
The investor selects specific securities or assets that are expected to decrease in value over time.
Model
Timing

3. Timing

Deciding if being invested or not (cash) on a basket of equities.
Customization
Model
Picking

3. Timing User Build

Choose Model & Tune Parameters

RNN
ARMA
GARCH
MLP
  • N
  • R
  • N
  • Number of Hidden Layers
  • Number of Neurons in Hidden Layers
  • Loss
  • Optimizer
  • Regularizator
The above parameters are only some of the many possible examples

3. Equity's Investment Strategy

Input
Performance Evaluation
Models
Target
Validation
M1
M5
M4
M3

M2

3. Model User Build

Model choice & Hyperparameters tuning

Linear
Random Forest
MLP
  • Regularization
  • Robust Variance
  • Autocorrelation Adjustment
  • Number of Trees
  • Bagging / Boosting
  • Learning Rate
  • Number of Hidden Layers
  • Number of Neurons in Hidden Layers
  • Loss
  • Optimizer
  • Regularization
The above parameters are only some of the many possible examples

3. Custom Optimization

Portfolio factor tilt
Build a portfolio that has a specific exposure towards some predefined factors, preserving as possible the allocation defined by a ranking (Qi4M's ranking from stock picking model or a ranking provided by the User).
Portfolio Turnover
Build a portfolio with a maximum level of turnover preserving as possible the allocation defined by a ranking (Qi4M's ranking from stock picking model or a ranking provided by the User).

3. Input

Financial Statement data
Premade Feature Templates
Sentiment data
Technical indicators
Custom variables
Macroeconomic data
  • Frequency (weekly/monthly/ quarterly)
  • Aggregation (12 months/quarterly)
Insights from financial articles, news, and social media, shaping our understanding of market movements.
  • Frequency (depending on the series: daily-quarterly)
  • Aggregation (12 months/quarterly)
  • Type of indicator
  • Lenght of lookback window
  • Qi4M Template A
  • Qi4M Template B
  • ...
  • DataTemplate X
  • DataTemplate Y
  • ...
User defined custom variables starting from raw Fiancial statement, Sentiment data and Technical indicators
  • Var. Transformations/Interactions
  • Dummy/Binary/Ordinal Encoding
  • Series denoising or decomposition
  • Series differentiation
Feature Engineering
DataTemplate

3. Target

Returns
Sharpe Ratio
Volatility
Info Ratio
Customize

3. Validation

Model Validation

Statistical Metrics

Financial Evaluations

3. Model Validation

Walk Forward Validation

Cross Validation

Train/Test Split

It consists of dividing the dataset in In-sample data IS (for model creation and tuning) and Out-of-sample data OOS (for model's performance validation). The IS set is used to find and optimize the model parameters, while the OOS set is used to evaluate the performances in a realistic setting using unseen data.

The dataset is divided into k-folds of equal size. The model is then tuned on k-1 folds and validated on the remaining fold. This process is repeated multiple times for robustness. The final outcome is obtained by aggregating the OOS results for each timestamp in the different iterations.

The model is tuned using IS data up to a certain point in time, and then it is evaluted on the subsequent unseen OOS timestamp.WF can be performed with: (a) expanding window: IS data window is increased over time, always maximising the amount of information available; (b) rolling window: IS data window is fixed and over time older obs. are discarded.

Out-of-sample data
In-sample data
Out-of-Sample Data

3. Statistical Metrics

Confusion Matrix: to check the model ability to classify correctly items in different classes, monitoring Type-I and Type-II errors.

Shapley Values: checking what the most important features are for the model's predictions and the impact each feature has on the final outcome.

Loss convergence: monitoring the evolution of the loss function in various epochs to check if the model properly converges during training.

4. Performance Evaluation

performance metrics vs. Benchmark
Performance metrics
Excess Return
NAV
Sharpe
Sortino
Tracking Error
or
Hit Ratio
Info Ratio
Alpha
WinLoss
Beta
Sample Evaluation

4. Performance Evaluation

2. Equity Index/ETFs Universe

Equity Index examples
  • S&P 500
  • Dow Jones Industrial Average (DJIA)
  • NASDAQ Composite
  • FTSE 100
  • Nikkei 225
  • Euro Stoxx 50
  • DAX

Custom

2. Custom Index

Sector
Geographical Area
Custom Feature via Qi4M Algortithm (1, 2, 3...)
Market Filters
Metaverse
AI
Sport
Automotive
Hydrogen
Green Food
Travel
Customize
Fashion
Green Food
Travel
Energy
Health Care
Financials
Industrials
Utilities
Technology
Technology
Materials
IT
ESG score
Market Cap
Volume
P/E Ratio
Dividend Yield
Regional
National
Global
Index

3. Equity Index/ETFs Investment Strategy

Time Horizons

Algorithmic Trading
Daily
Weekly
Timing
Monthly
Quarterly

3. Algorithmic Trading

Momentum/Trend Following
Focus on capitalizing on prevailing trends by entering or exiting positions based on the momentum or strength of price movements in the market.
Mean Reversal
Short
Taking positions based on the expectation that prices will revert to their historical average after experiencing significant deviations, aiming to profit from the correction.
Statistical Arbitrage Strategy
Seeks to exploit relative price movements between related assets by identifying and capitalizing on statistically significant deviations from their historical relationships or correlations.
Model
Timing

3. Equity Index/ETFs Investment Strategy

Timing

Cash/Saving

Automated trading strategy based that analyze market data and financial indicators to determine the ideal time to enter and exit financial markets, dynamically allocating AUM between cash and the risky asset.

Model
Algorithmic Trading

3. Equity Index/ETFs Investment Strategy

Input
Performance Evaluation
Model
Target
Validation

3. Model User Build

User Build: Choose Model & Tune Parameters

Linear
Random Forest
MLP
  • Regularization
  • Robust Ear
  • Autocorelation Adjustment (?)
  • Number of Trees
  • Bagging / Roosting
  • Learning Rate
  • Number of Hidden Layers
  • Number of Neurons in Hidden Layers
  • Loss
  • Optimizer
  • Regularizator
The above parameters are only some of the many possible examples

3. Target

Returns
Sharpe Ratio
Volatility
Customize

3. Validation

Model Validation

Statistical Metrics

Financial Evaluations

3. Model Validation

Walk Forward Validation

Cross Validation

Train/Test Split

It consists of dividing the dataset in In-sample data IS (for model creation and tuning) and Out-of-sample data OOS (for model's performance validation). The IS set is used to find and optimize the model parameters, while the OOS set is used to evaluate the performances in a realistic setting using unseen data.

The dataset is divided into k-folds of equal size. The model is then tuned on k-1 folds and validated on the remaining fold. This process is repeated multiple times for robustness. The final outcome is obtained by aggregating the OOS results for each timestamp in the different iterations.

The model is tuned using IS data up to a certain point in time, and then it is evaluted on the subsequent unseen OOS timestamp.WF can be performed with: (a) expanding window: IS data window is increased over time, always maximising the amount of information available; (b) rolling window: IS data window is fixed and over time older obs. are discarded.

Out-of-sample data
In-sample data
Out-of-Sample Data

3. Statistical Metrics

Confusion Matrix: to check the model ability to classify correctly items in different classes, monitoring Type-I and Type-II errors.

Shapley Values: checking what the most important features are for the model's predictions and the impact each feature has on the final outcome.

Loss convergence: monitoring the evolution of the loss function in various epochs to check if the model properly converges during training.

3. Input

Premade Feature Templates
Macroeconomic data
Techincal indicators
  • Frequency (depending on the series: daily-quarterly)
  • Aggregation (12 months/quarterly)
  • Type of indicator
  • Type of lookback wdw
  • Qi4M Template A
  • Qi4M Template B
  • ...
  • DataTemplate X
  • DataTemplate Y
  • ...
  • Variables Transformations
  • Encoding
  • Ordinal Encoding
  • Series denoising or deconmposition
  • Series differentiation
Feature Engineering
DataTemplate

4. Performance Evaluation

performance metrics vs. Benchmark
performance Metrics
Excess Return
NAV
Sharpe
Sortino
CAGR
Tracking Error
or
Hit Ratio
Info Ratio
Alpha
WinLoss
Beta

2. Commodities Universe

Commodities examples
  • Oil
  • Gas
  • Energy
  • Cotton
  • Sugar
  • Coffee
  • Cocoa
  • Gold
  • Silver
  • Copper
  • Aluminum
  • Wheat
  • Corn

3. Commodities Investment Strategy

Time Horizons

Algorithmic Trading
Daily
Weekly
Monthly
Timing
Quarterly

3. Algorithmic Trading

Algorithmic Trading

Momentum/Trend Following
Focus on capitalizing on prevailing trends by entering or exiting positions based on the momentum or strength of price movements in the market.
Mean Reversal
Taking positions based on the expectation that prices will revert to their historical average after experiencing significant deviations, aiming to profit from the correction.
Statistical Arbitrage Strategy
Seeks to exploit relative price movements between related assets by identifying and capitalizing on statistically significant deviations from their historical relationships or correlations.
Model
Timing

3. Timing

Timing

Cash/Saving

Automated trading strategy based that analyze market data and financial indicators to determine the ideal time to enter and exit financial markets, dynamically allocating AUM between cash and the risky asset.

User Build
Model
Algorithmic Trading

3. Timing User Build

Timing User Build: Choose Model & Tune Parameters

RNN
ARMA
GARCHI
MLP
  • N
  • R
  • N
  • Number of Hidden Layers
  • Number of Neurons in Hidden Layers
  • Loss
  • Optimizer
  • Regularizator
The above parameters are only some of the many possible examples

3. Commodities Investment Strategy

Input
Performance Evaluation
Model
Target
Validation

3. Model User Build

Choose Model & Tune Parameters

Linear
Random Forest
MLP
  • Regularization
  • Robust Variance
  • Autocorrelation Adjustment
  • Number of Trees
  • Bagging / Boosting
  • Learning Rate
  • Number of Hidden Layers
  • Number of Neurons in Hidden Layers
  • Loss
  • Optimizer
  • Regularization
The above parameters are only some of the many possible examples

3. Input

Premade Feature Templates
Techincal indicators
Macroeconomic data
  • Frequency (depending on the series: daily-quarterly)
  • Aggregation (12 months/quarterly)
  • Type of indicator
  • Type of lookback wdw
  • Qi4M Template A
  • Qi4M Template B
  • ...
  • DataTemplate X
  • DataTemplate Y
  • ...
  • Variables Transformations
  • Encoding
  • Ordinal Encoding
  • Series denoising or deconmposition
  • Series differentiation
Feature Engineering
DataTemplate

3. Target

Target

Returns

Sharpe Ratio

Volatility

Custom

3. Validation

Model Validation

Statistical Metrics

Financial Evaluations

3. Model Validation

Walk Forward Validation

Cross Validation

Train/Test Split

It consists of dividing the dataset in In-sample data IS (for model creation and tuning) and Out-of-sample data OOS (for model's performance validation). The IS set is used to find and optimize the model parameters, while the OOS set is used to evaluate the performances in a realistic setting using unseen data.

The dataset is divided into k-folds of equal size. The model is then tuned on k-1 folds and validated on the remaining fold. This process is repeated multiple times for robustness. The final outcome is obtained by aggregating the OOS results for each timestamp in the different iterations.

The model is tuned using IS data up to a certain point in time, and then it is evaluted on the subsequent unseen OOS timestamp.WF can be performed with: (a) expanding window: IS data window is increased over time, always maximising the amount of information available; (b) rolling window: IS data window is fixed and over time older obs. are discarded.

Out-of-sample data
In-sample data
Out-of-Sample Data

3. Statistical Metrics

Confusion Matrix: to check the model ability to classify correctly items in different classes, monitoring Type-I and Type-II errors.

Shapley Values: checking what the most important features are for the model's predictions and the impact each feature has on the final outcome.

Loss convergence: monitoring the evolution of the loss function in various epochs to check if the model properly converges during training.

4. Performance Evaluation

performance metrics vs. Benchmark
performance Metrics
Excess Return
NAV
Sharpe
Sortino
CAGR
Tracking Error
or
Hit Ratio
Info Ratio
Alpha
WinLoss
Beta

2. Crypto Universe

Crypto examples
  • Bitcoin
  • Ethereum

2. Fixed Income Universe

Sector
Basket Universe from Market Index
Geographical Area
Filters
Energy
Health Care
Financials
Industrials
Utilities
Technology
Technology
Materials
IT
Emission Size
Currency
Duration
Seniority
Rating
ESG Score
Minimum YTM
EU HY Corp.
EU IG Corp.
US IG Corp.
EU Govt.
US HY Corp.
Regional
National
Global
Universe

3. Fixed Income Investment Strategy

WEIGHTING

Picking

Time Horizons

Custom Constraints
Longlist Selection: produces a ranking based on the bonds' characteristics.
Duration
Seniority
Rating
Sector Allocation
Geo/Region
Currency
Emission Size
Liquidation
ESG Score
Min YTM
Monthly
Quarterly
Yearly

3. Fixed Income Investment Strategy

Input
Performance Evaluation
Model
Target
Validation

3. Input

Bond Characteristics / Pricing Data
Premade Feature Templates
Macroeconomic data
Sentiment data
Techincal indicators
  • Frequency (depending on the series: daily-quarterly)
  • Aggregation (12 months/quarterly)
  • Frequency (weekly/monthly/ quarterly)
  • Aggregation (12 months/quarterly)
  • Type of indicator
  • Type of lookback wdw
  • On firms
  • On broader economic conditions
  • Qi4M Template A
  • Qi4M Template B
  • ...
  • DataTemplate X
  • DataTemplate Y
  • ...
  • Variables Transformations
  • Encoding
  • Ordinal Encoding
  • Series denoising or deconmposition
  • Series differentiation
Feature Engineering
DataTemplate

3. Target

Target

Returns

Sharpe Ratio

Volatility

Custom

3. Validation

Model Validation

Statistical Metrics

Financial Evaluations

3. Model User Build

Choose Model & Tune Parameters

Linear
Random Forest
MLP
  • Regularization
  • Robust Variance
  • Autocorrelation Adjustment
  • Number of Trees
  • Bagging / Boosting
  • Learning Rate
  • Number of Hidden Layers
  • Number of Neurons in Hidden Layers
  • Loss
  • Optimizer
  • Regularization
The above parameters are only some of the many possible examples

3. Model Validation

Walk Forward Validation

Cross Validation

Train/Test Split

It consists of dividing the dataset in In-sample data IS (for model creation and tuning) and Out-of-sample data OOS (for model's performance validation). The IS set is used to find and optimize the model parameters, while the OOS set is used to evaluate the performances in a realistic setting using unseen data.

The dataset is divided into k-folds of equal size. The model is then tuned on k-1 folds and validated on the remaining fold. This process is repeated multiple times for robustness. The final outcome is obtained by aggregating the OOS results for each timestamp in the different iterations.

The model is tuned using IS data up to a certain point in time, and then it is evaluted on the subsequent unseen OOS timestamp.WF can be performed with: (a) expanding window: IS data window is increased over time, always maximising the amount of information available; (b) rolling window: IS data window is fixed and over time older obs. are discarded.

Out-of-sample data
In-sample data
Out-of-Sample Data

3. Statistical Metrics

Confusion Matrix: to check the model ability to classify correctly items in different classes, monitoring Type-I and Type-II errors.

Shapley Values: checking what the most important features are for the model's predictions and the impact each feature has on the final outcome.

Loss convergence: monitoring the evolution of the loss function in various epochs to check if the model properly converges during training.

4. Performance Evaluation

performance metrics vs. Benchmark
Performance metrics
Excess Return
NAV
Sharpe
Sortino
CAGR
Tracking Error
or
Hit Ratio
Info Ratio
Alpha
WinLoss
Beta