case-based recommender systems for personalized finance advisory

Cataldo Musto, Giovanni Semeraro

Case-based Recommender Systems for Personalized Finance Advisory

Graz (Austria) - 16.04.2015

one minute on the Web

C.Musto, G.Semeraro - Case-based Recommender Systems for Personalized Finance Advisory FinRec 2015 - 1st International Workshop on Personalization and Recommender Systems in Financial Services - Graz (Austria) - 16.04.2015

we can handle 126 bits of information we deal with 393 bits of information

ratio: more than 3x(Source: Adrian C.Ott, The 24-hour customer)

(from Matrix)

decision-making is actually challenging

paradox of choice(Barry Schwartz, TED talk “Why more is less”)

(financial) overloadC.Musto, G.Semeraro - Case-based Recommender Systems for Personalized Finance Advisory FinRec 2015 - 1st International Workshop on Personalization and Recommender Systems in Financial Services - Graz (Austria) - 16.04.2015

solution: personalizationC.Musto, G.Semeraro - Case-based Recommender Systems for Personalized Finance Advisory FinRec 2015 - 1st International Workshop on Personalization and Recommender Systems in Financial Services - Graz (Austria) - 16.04.2015

to adapt asset portfolios

on the ground of personal user profile and needs

Insight:

SolutionRecommender Systems

Recommender Systems

Relevant items (movies, news, books, etc.) are suggested to the user according to her preferences.

definitionRecommender Systems have the goal of guiding the

users in a personalized way to interesting

or useful objects in a large space of possible options.

Burke, 2002 (*)(*) Robin D. Burke: Hybrid Recommender Systems: Survey and Experiments. UMUAI, volume 12, issue 4, 331-370 (2002)

does it fit our scenario?“we are leaving the age of information, we are entering the age of recommendation”

(C.Anderson, The Long Tail. Wired. October 2004)

Recommender Systems

“[...] The technology is used by shopping websites such as Amazon, which receives about 35 percent of its revenue via product recommendations. It is also used by coupon sites like Groupon; by travel sites to suggest flights, hotels, and rental cars; by social-networking sites such as LinkedIn; by video sites like Netflix to recommend movies and TV shows, and by music, news, and food sites to suggest songs, news stories, and restaurants, respectively. Even financial-services firms recently began using recommender systems to provide alerts for investors about key market events in which they might be interested”

(N.Leavitt, “A technology that comes highly recommended” - http://tinyurl.com/d5y5hyl)

Recommender Systemsfinancial services

http://www.bloomberg.com/news/articles/2015-03-16/smart-beta-etfs-attract-billions-with-critics-blaming-dumb-money

Recommender Systemssuccess stories

“People who bought…”on Amazon

“Discover”on Spotify

Recommender SystemsRecommender Systemsunexpected stories

recommending financial products is a complex task

flocking

Too many users could be moved towards the same suggestions

flocking

consequence: price manipulation (as in trader forums)

poor knowledge

Features describing both assets classes and private investors are

poorly meaningful

poor knowledge

poor history

A combination of asset classes is typically kept for a long time

Solution

Case-based Recommender SystemsC.Musto, G.Semeraro - Case-based Recommender Systems for Personalized Finance Advisory FinRec 2015 - 1st International Workshop on Personalization and Recommender Systems in Financial Services - Graz (Austria) - 16.04.2015

case-based RSs• Inspired by case-based reasoning

• Similar problems solved in the past are used as knowledge base

• Reasoning by analogy

• The recommendation process relies on the retrieval and the adaptation of the solutions adopted to solve similar cases

....butwhat do we actually mean with ‘case’ ?

case base

• A case is a the formalization of a previously solved problem

• In our setting

• Description of a user

• Description of a portfolio

• An evaluation of the proposed solution

case-baseexample

user solution evaluation

case-baseexample

User Features Risk Profile: LowFinancial Experience: HighFinancial Situation: Very HighInvestment Goals: MediumTemporal Goals: Medium

case-baseexample

Euro Bonds 30%

High-Yield Bonds 10%

Fixed-Rate bonds 22%

Euro Stocks 23%

Emerging Market Stocks 7%

Money Market 8%

case-baseexample

monthly rate (e.g.)

+0.22%

Euro Bond 30%

High-Yield Bonds 10%

Fixed-Rate bonds 22%

Euro Stocks 23%

Emerging Markets Stocks 7%

Money Market 8%

case-based RSssolving cycle

case-based reasoning for personalized wealth management

scenario

“Scrooge McDuck wants to get richer. He decided to invest some of his savings and he asked for help to a

financial advisor”

step 1 user modeling

scenario

Which features may describe

Scrooge McDuck?

step 1

scenario

User Features Risk Profile: Low

Investment Horizon HighInvestment Experience Very High

Investment Goals: MediumFinancial Assets: Medium

step 1

scenario

MiFID-based

step 1

scenariostep 1

+Generic Demographical Features

in a classical pipeline, the target user

would have received a “model” portfolio tailored on her profile

in a pipeline fostered by a recommender system, the financial advisor can analyze the portfolios proposed to similar users

to tailor the proposal

step 2 neighbors identification

given a case base, it is necessary to

define a similarity measure to compute how similar two cases are

neighbors identificationtrivial similarity: user match

two cases are similar if they share exactly the same features

trivial similarity: user match

two cases are similar if they share exactly the same features

neighbors identification

cases are represented as points in a vector space

geometrical alternative: cosine similarity

geometrical representationC.Musto, G.Semeraro - Case-based Recommender Systems for Personalized Finance Advisory FinRec 2015 - 1st International Workshop on Personalization and Recommender Systems in Financial Services - Graz (Austria) - 16.04.2015

geometrical alternative: cosine similarityneighbors identification

case-based RSsgeometrical alternative: cosine similarity

each case is seen as a vector

calculation over the n features

= (risk profile, experience, goals, etc.)

inner product

it returns the cosine of the angle between A and B

case_A

case_B

cosine

scenario

case base

step 2

scenariostep 2

similarity score

scenariostep 2

neighborhood(helpful cases)

step 3 extraction of candidate portfolios

scenario

Euro Bonds 30%High Yield Bonds 15%Fixed Rate Bonds 15%

Europe Stocks 20%Emerging Markets Stocks 12%

Money Market 8%

Euro Bonds 30%High Yield Bonds 10%Fixed Rate Bonds 22%

Europe Stocks 23%Emerging Markets Stocks 7%Flessibili Bassa Volatilità 8%

step 2

solutions proposed to the neighbors are labeled as candidate solutions

step 4 ranking of candidate portfolios

in real-world scenarios, the case base

contains many helpful cases

in real-world scenarios, the case base

contains many helpful cases

it is necessary to introduce strategies to filter and rank the cases

revise

We implemented several ranking strategies

• Temporal ranking

• Clustering

• Diversification

• Financial Confidence Value (FCV)

revisetemporal ranking

solutions are ranked from the newest to the oldest (or viceversa)

Euro Bonds 30%

High Yield Bonds 15%

Fixed Rate Bonds 15%

Europe Stocks 20%

Money Market 8%

Euro Bonds 30%

Europe Stocks 23%

Money Market 8%

Euro Bonds 15%

Europe Stocks 40%

Money Market 8%

Euro Bonds 20%

Europe Stocks 35%

Money Market 8%

Euro Bonds 30%

Europe Stocks 20%

Money Market 8%

olderolder

Euro Bonds 30%

Europe Stocks 23%

Money Market 8%

Euro Bonds 15%

Europe Stocks 40%

Money Market 8%

Euro Bonds 20%

Europe Stocks 35%

Money Market 8%

clustering

solutions are clustered and just a small set of centroids is proposed

revise

clusteringrevise

Euro Bonds 30%

Europe Stocks 20%

Money Market 8%

Euro Bonds 30%

Europe Stocks 23%

Money Market 8%

Euro Bonds 15%

Europe Stocks 40%

Money Market 8%

Euro Bonds 20%

Europe Stocks 35%

Money Market 8%

clusteringcluster 1

reviseEuro Bonds 30%

Europe Stocks 20%

Money Market 8%

Euro Bonds 30%

Europe Stocks 23%

Money Market 8%

Euro Bonds 15%

Europe Stocks 40%

Money Market 8%

Euro Bonds 20%

Europe Stocks 35%

Money Market 8%

clusteringcluster 1 cluster 2

reviseEuro Bonds 30%

Europe Stocks 20%

Money Market 8%

Euro Bonds 30%

Europe Stocks 23%

Money Market 8%

Euro Bonds 15%

Europe Stocks 40%

Money Market 8%

Euro Bonds 20%

Europe Stocks 35%

Money Market 8%

insight: filtering out too similar solutions

diversification algorithmrevise

revise

identification of the best subset of similar cases which maximize the relative diversity

diversification algorithm

revisediversification algorithm

revise

input similar cases

(candidate solutions)

revise

output subset of

diversified cases

revise

algorithm in each step the portfolio which best diversifies the solutions is

chosen

revise

Solutions with the highest quality are iteratively

chosen

revise

combination between

similarity and diversity

it returns portfolios that are not so similar to those

previously put in the result set

Euro Bonds 30%

Europe Stocks 20%

Money Market 8%

Euro Bonds 30%

Europe Stocks 23%

Money Market 8%

Euro Bonds 15%

Europe Stocks 40%

Money Market 8%

Euro Bonds 20%

Europe Stocks 35%

Money Market 8%

Euro Bonds 30%

Europe Stocks 20%

Money Market 8%

Euro Bonds 30%

Europe Stocks 23%

Money Market 8%

Euro Bonds 15%

Europe Stocks 40%

Money Market 8%

Euro Bonds 20%

Europe Stocks 35%

Money Market 8% XXC.Musto, G.Semeraro - Case-based Recommender Systems for Personalized Finance Advisory FinRec 2015 - 1st International Workshop on Personalization and Recommender Systems in Financial Services - Graz (Austria) - 16.04.2015

reviseFinancial Confidence Value (FCV)

• Simple insight

• We know the historical yield for each of the assets class in the portfolio

• FCV ranks first the solutions composed by a combination of asset classes close to the optimal one (according to previous yield)

revise

(Generated yield) (Drift Factor)Total yield is the product of the

yield generated by each asset

class with the its percentage in the

portfolio

Ratio between the yield

generated by the asset classes in the portfolio and its complement

Financial Confidence Value (FCV)

revise

Euro Bonds - - - 30%

Europe Stocks +++ 20%

Money Market 8%

revise

Money Market 8%

step 5 discussion of the solution

financial advisor and private investor

can further discuss the portfolio

review

Original Discussed Gap

Euro Bonds 30% 30%High Yield Bonds 12.5% 10% -2.5%Fixed Rate Bonds 18.5% 20% +1.5%

Europe Stocks 21.5% 24% +2.5%Emerging Markets

Stocks 9.5% 8% -1.5%Money Market 8% 8%

interactive personalization

step 6 case base update

an evaluation score is finally assigned to the proposed solution

yield, e.g.

retain

good solutions are stored in the case base and exploited for future recommendations

case base

(new) case base

our implementation

our implementationhttp://193.204.187.192:8080/OBWFinance

demo available

OBWFinancelogin screen

advisor-oriented tool

OBWFinanceclient selection

OBWFinancerecommendation parameters

OBWFinanceonly admins can change the parameters

OBWFinanceone click to generate recommendations

OBWFinancedrop-down menu for selecting the best solution

OBWFinanceassets class

OBWFinanceyield of the solution

OBWFinancechosen portfolio can be further discussed

evaluation

evaluationwhat is the average yield of

recommended portfolios?

evaluationwhat is the average yield of

recommended portfolios?

can recommender systems suggest

better investment portfolios than human advisors?

design of the experiment

• 1172 users

• 19 assets classes

• Different neighborhood sizes

• Different features describing the users

• Risk Profile, Investment Goals, Investment Horizon, Investment Experience, Financial Assets, Advice Type, Sex, Age

• Different similarity measures (Cosine vs. UserMatch)

• Leave-one-out experimental design

evaluation

experiment 1user match vs. cosine similarity

neighbors

1 5 10

0,20,190,18

0,10,110,09

User Match Cosine Sim

cosine similarity overcomes user match

experiment 2how many features?

neighbors

1 5 10

0,20,210,2 0,20,190,18

Financial Features Financial + Demographical Features

cosine similarity overcomes user match

experiment 3revise strategies (yield)

best performing configuration provides 0,28% monthly yield

neighbors

1 5 10

0,250,240,22

0,270,28

0,220,2

0,150,13 0,14

0,120,09

0,20,210,2

Basic Clustering Diversification FCV FCV + Div

experiment 3revise strategies (diversity of the solutions)

ILD=1-average similarity between portfolios

neighbors

0,460,41

Basic Clustering Diversification FCV FCV + Div

experiment 4comparison to baselines (leave-one-out evaluation)

recsys better than humans!

neighbors

1 5 10

0,270,28

0,220,20,20,2

0,170,170,17

Human Collaborative FCV

• FCV calculated on January, 2014

• Recommendations generated on January, 2014

• Evaluation of the yield generated from February 2014 to July 2014

experiment 5ex-post evaluation (6 months, with real data)

FCV and Diversification is the best one

neighbors

1 5 10

0,060,060,060,040,04

0,110,12

0,090,1

0,060,08

Basic FCV FCV + Div Collaborative Human

• Personalized Wealth Management

• Application of case-based reasoning

• Geometrical similarity measure to identify the most similar previously solved cases

• Introduction of diversification and re-ranking techniques

• More than 3% yield for year

• Experiments shows that recommended portfolios overcome the real ones for almost all the users

• Working Demo!

questions?

Giovanni Semeraro giovanni.semeraro@uniba.it

Cataldo Musto cataldo.musto@uniba.it

in memoriam

Aaron Swartz

case-based recommender systems for personalized finance advisory

hybrid recommender systems

definitionrecommender

semeraro case

international workshop

personalized way

cataldo musto

financial overloadc

adrian c

Technology

blending human computing and recommender systems for...

personalized recommender systems: mining user...

personalized links recommendation based on data mining in...

personalized news recommender using twitter

recommender tv a personalized tv guide system compliant...

personalized recommender system on whom to follow in twitter

personalized recommender system using entropy based...

personalized wealth management through case-based...

personalized recommendation via cross-domain triadic ... ·...

deep learning for personalized search and recommender...

mubs: a personalized recommender system for behavioral

a recommender model for the personalized adaptive chunk

personalized movie recommender

recommender systems for mass customization of financial...

research on personalized recommender system for tourism...

purs: personalized unexpected recommender system for … ·...

an evaluation of simrank and personalized pagerank to build...

personalized explanations for hybrid recommender...

hybrid recommender system architecture for personalized...

personalized explanations for hybrid recommender...