2 Assignment - 200 + Words EACH - MIS Video Assignment - 12 Hours Maximum

Chapter 13

Expert Systems in Baltimore County Police Department

Can Crime Prediction Software Stop Criminals? http://www.youtube.com/watch?v=s1-pbJKA3H8

New crime prediction software called PredPol (short for predictive policing) aims to make police work anything but random. PredPol identifies areas in your division that are at the highest risk of a crime occurring during a certain time period. The Foothill community police station, a division of the LA police department, is one of the first police departments in the United States to test PredPol. Twice a day Foothill police officers get a list of twenty five hundred foot by five hundred foot boxes to patrol. The most common crimes are burglary for motor vehicle and grand theft auto. The results of the software are promising; for example, burglary had a 25% decrease. PredPol identifies crime generators: places full of people or cars that create opportunities for criminals. PredPol also forcasts repeat victimization; for example, if your house was broken into once, there is a higher chance that it will happen again. PredPol also looks at near repeat victimization meaning if your neighbor’s house was burglarized then you also have higher odds that your house will be burglarized too. PredPol does not predict who will commit a crime only where a crime will occur. It may actually reduce the number of arrests by stopping crimes before they start. Foothill Police Department’s Captain estimates that PredPol has helped saved about $2 million in his division alone by helping to reduce crime.

Intelligent Agents in Action

Intelligent Virtual Call Agents: The Next Generation of Voice Self-Service http://www.youtube.com/watch?v=CLEtJJqcZHk

VPI Virtual Source is a revolutionary hosted pay as you go service that uses virtual call agents powered by artificial intelligence. With Virtual Source you can now automate a much broader range of call types for just one fifth of the cost of a human agent. Virtual Source also allows you to achieve higher call completion and customer satisfaction rates than using traditional speech IVRs or human agents. VPI Virtual Source allows for virtual agents to automate a wide variety of your inbound and outbound call types and offload the mundane, repetitive tasks that waste the time of skilled agents and needlessly drive up costs. Virtual agents use free flow conversation, are super intelligent, and always pleasant to talk to. They acquire their knowledge and skills by training. They supplement and leverage your existing systems and they are available 24/7. Virtual agents are a great conversational front door to gather and verify information upfront for your human agents.

Fuzzy Logic in Action

An Egg-Boiling Fuzzy Logic Robot http://www.youtube.com/watch?v=J_Q5X0nTmrA

Fuzzy logic is a tool that engineers and scientists use to add intelligence into robots, video games, and household appliances. Fuzzy logic solves complex problems by using simple rules. For example, if we want to teach a robot how to boil an egg, three steps need to be taken. In step 1, fuzzification occurs where the robot considers an egg to be either small or large but in different degrees, which are called fuzzy values. In step 2, inference occurs where the robot applies the rules it has been given in different degrees. In step 3, defuzzification occurs where the robot must decide the boiling time based on the fuzzy values where a balanced decision is made. Fuzzy logic is used in camera stabilization systems, washing machines, and automobile braking systems among others.

Neural Networks in Microsoft and the Chicago Police Department BOX

AANN - Legend's Asset Allocation Neural Network

http://www.youtube.com/watch?v=OsEediXqU94

AANN is a neural network that’s been known as the asset allocation neural network and was developed to assist in allocating assets within mutual fund portfolios. Neural networks can synthesize huge amounts of information without fatigue and respond without emotion or bias. Neural networks are adaptive systems that can actually learn from their experiences. At the start of AANN’s training years ago, AANN was instilled with one specific goal: to determine the relative strength of each asset class within a given set. AANN was given a string of variable that can impact world investment markets. AANN examined and reexamined these inputs countless times until AANN came up with the correct response for each pattern of inputs. Through the process of training and testing, AANN teaches itself to recognize complex relationships among a large number of constantly changing variables. Once these correlations emerge, AANN can predict the impact these variables will have on each asset class during a specific period of time. AANN’s output provides rankings for 7 asset classes according to their perceived potential to generate positive returns. AANN’s recommendations are then entered into a quantitative optimization modeling program, which determines an appropriate mix of investments for a specific mutual fund portfolio given its objective and tolerance for risk.