Future Work and Conclusion
From Kb
Chapter 6 concludes with a review of the thesis and outlines possible improvements to make to the framework and makes suggestions for future research.
Contact Article Author | Blog of Article Author | FirstPartners.net Home | LinkedIn profile of Author
Dissertation Future work
1 Chapter 6 – Conclusion and future work Todo: - Move bits 'below line' into format - push to first draft
1. Summary of where we have go to so far 2. Future work 1. Joone 2. Agents 3. Web 2 frameworks 1. Jruby 2. PHP 4. Other 1. Expand on existing ... 1. Fitnesse Testing 2. Upgrade of components 3. Other items
Nice to have - Jruby Integration - Fitnesse testing - Java business process JBPM - move to lucene 1.9 / 2.0 - PHP - Joone - Agents
Later use seam to integration workflow and rules
NTH: Drools Articles code as part of RP Snippets
NTH: FTC code samples as part of RP Snippets
NTH: Fitnesse Tests on rules
Future work Talk about (no tech sample)
TALK ABOUT HOW TO TIE THIS INTO WEB 2 - additional workflow / jbpel / oracle - additional rule engines (cover search bases) - enough to talk through how to extend search to use hibernate - bring in / reference stuff from 2 drools articles - bring in / reference stuff from struts artticles - go through articles (FTC) - put enough in the related technologies to CYA on those - Security - bit about assertions / design by contract - talk about hibernate mappings - bit on security (quite a bit on security – at least one page – Ageci Security) bit about neural networks
Future Work - Agents
People work be negociation , with no central authority. Why shouldn't your information act the same way?
More on this if we decide to make this part of the Web 2.0 spiel
Summary
Business Problem: Where several people have parts of the solution, but need to collaborate the get the answer Solution: Agents acting as a proxy for each person to collaborate in a market based env. Link to Web 2.0: Web 2.0 is effectively an agent environment (multithreaded systems interacting with each other over the web) Business Example: Sales / Prospect distribution
Business Problem: · Where several people have parts of the solution, but need to collaborate the get the answer
b) Solution:
· Agents acting as a proxy for each person to collaborate in a market based env.
c) Link to Web 2.0:
· Web 2.0 is effectively an agent environment (multithreaded systems interacting with each other over the web)
d) Business Example:
· Sales / Prospect distribution · 'Use Agents to plan your next Web 2.0 Unconference' · e.g. Resource Scheduling (for Education, for room booking, for travel) · Auction (Sellers and Buyers) · Marketplace (e.g. Brokers) · Sales profiling / bidding for leads · Time scheduling (what work on during week) ·
e) Technical Implementation
· Pseudo Agents using workflow · Why done this way · - Enterprise Java · - Tie back to agents · - Use workflow as round-robin · - Use scheduler! · - Auction / Trading Site for Sales Leads · Collective intelligence · idea of capitalism v communisum using baker example · Objects with Attitude (TM) · Rule engines (to show logic) · Web 2.0 messgaes ideas · Use IBM agent frameworks (or Cougaar) · Wisdom of crowds – setting prices · Deciding on which customers to 'own' · Can tie this into data? (e.g · Red Swarm (Financial Sales Edition) or 'Sales Profiler' · Intelligence from rule engine? (multiple or single filtered sheet) · - Use where there is a scarce resource (room , time , sales leads , capital, etc) that has to be shared between many people. Each person has some knowledge towards the solution (e.g their own pref , how highly they value a lead). · Allow external factors to set their prefernces · e.g. People for conference · Game theory comes from people (expect them to learn) · Internal rules (e.g. Money) cannot be hacked · · Use pseudo (or actual paypal) money to measure utility · Can set this money at the start (our way to influence system) · allows trading · allows measure of utility · stop simulation after x amount of time · either number of trades decreases / arbitaory : solution is better than before (may be more optimal , but 'good enough') · Start with capital of last months sales · Allow for length of client relationship · Bidding gives sales target · Random allocation of clients · agent for each sales rep · These get all · agent for 'rubbish' clients · buyer of last resort – injects liquidity · or don't use this, but everybody has to bid?? · All agents can see – make bid on anothers client · 2 Steps each cycle · Identify own item offer for bid · Identify other item offer to sales · Need to remember which items were bid for before – and where stop repititve se Auction to sell? (and if so , what is the best way – e.g. Sealed , one shot bid?) · Can be open and show logs to demonstrate fairness · How demo as flash or Web app? · System will need to run continuosly (be hosted? · Explain why this approach is better · allows users to set preferences · negociates / finds optimal solutions based on these preferences · can be extended to allow external parties · monetary value: · Can be extended to allow for more complex behavour · Market based approach (decentralised) that happens to use agents · Have time to run full process (then update process as piece by piece) · Tie into rule engine (as smarts) & tie into workflow (for update)
Own Thoughts · How overcome chicken and egg problem – where you need agents created on both sides (or do an ebay – where ebay give you the agent to program) · Can we apply this to market rules (e.g. Agent representing a customer??) · Marketing · How does Cougar fit into this? · Agents are objects with personalities · Anything (Agents / General) that I can pick up from MIT? · Like C/ C++ (grafted OO onto it) – then Java replace. Current state is Java with Agents grafted onto it (will this evolve to a new language) · Objects are becomming more and more like agents anyway in repsonse to agents / web 2.0 (10 years ago we lived in static computer env) . Now most are dynamic · Objects tend to represent things, while agents tend to represent people · Trade off – do we we need agents ,or perhaps a multi-threaded algorithm · Not so much multiple agents running on the same server , but multiple servers interacting
Future Work – Joone
Do this in the style of the shorts (business problem etc.)
