- Talk with managers or use research to identify two expert systems (that are not described in the textbook). What tools were used to build the expert system? What problem is it designed to solve?
- An HR manager wants to develop an expert system to evaluate potential employees applying to work for a job in your department. Assuming the job can be performed by a business-school intern, list some of the questions you would ask potential employees. Create a decision tree to evaluate the basic questions. Try to generate three possible outcomes: acceptable, unacceptable, and personal interview to decide. Your goal is to reduce the number of people needing a personal interview.
- Interview an expert in some area and create an initial set of rules that you could use for an expert system. If you cannot find a cooperative expert, try researching one of the following topics in your library: fruit tree propagation and pruning (what trees are needed for cross-pollination, what varieties grow best in each region, what fertilizers are needed, when they should be pruned); requirements or qualifications for public assistance or some other governmental program (check government documents); legal requirements to determine whether a contract is in effect (check books on business law).
- Describe how you could use data mining tools to help you find a new vehicle to buy. How well do the car-buying sites such as Edmunds perform these tasks for you?
- Obtain an expert system (e.g., Jess and CLIPS are free). Create a set of rules to evaluate a simple request for a car loan.
- Identify a problem that would be well suited for a neural network. Explain how the system would be trained (e.g., what existing data can be used?). Explain why you think the problem needs a neural network and what benefits can be gained.
- For the following problems identify those that would be best suited for an expert system, decision support system, or a more advanced AI system. Explain why.
- Helping an online shopper choose accessories for her business suit.
- Helping a high-school student choose a university to attend.
- Evaluating a new product to determine the best marketing campaign, specifically to help allocate money between radio, television, print, and online ads.
- Deciding how much to pay the CEO.
- Pricing a consumer product that has high seasonal demands, such as soft drinks or candy.
- Determining a location for a new fast food restaurant.
- Forecasting the outcome of a national election.
- Who will pay for the creation of software agents? What about the use of the agents? Should (or could) users be charged every time their agent calls another one? What about network usage? What would happen if your agent used your telephone to connect to thousands of other agents?
- Use a spreadsheet to create the example from the Human Resources Management example. Fill in the market adjustment column so that raises match the performance appraisals. Remember, total raises cannot exceed $10,000.
You are a midlevel manager for a small department store. You have collected a large amount of data on sales for 2004. Your transaction system kept track of every sale (order) by customer. Most customers paid by credit card or check, so you have complete customer data. Walk-in customers who paid cash are given a separate customer number, so you still have the sales data.
You are trying to determine staffing levels for each department. You know that the store becomes much busier during the end-of-the-year holiday season. For summer months, you have thought about combining staff from the departments. From conversations with experienced workers, you have determined that there is a maximum number of customers that can be handled by one person in a department. These numbers are expressed as monthly averages in the table.
You are thinking about combining workers from some of the departments to save on staffing—especially over the spring and summer months. However, working multiple departments makes the sales staff less efficient. There are two considerations in combining staff members. First, if any of the departments are reduced to a staff of zero, sales in that department will drop by 10 percent for that month. Second, total staffing should be kept at the level defined by the monthly averages. If average staffing (total across all departments) falls below the total suggested, then sales in all departments will fall by 2 percent for each tenth of a percentage point below the suggested average.
| Department | Customers/month |
|---|
| Clothing-Children | 180 |
| Clothing-Men | 150 |
| Clothing-Women | 180 |
| Electronics | 200 |
| Furniture | 150 |
| Household | 250 |
| Linen | 300 |
| Shoes | 300 |
| Sports | 400 |
| Tools | 340 |
- Using the database and a spreadsheet, determine how many workers we need in each department for each month. Present a plan for combining departments if it can save the company money. Assume that sales members cost an average of $1,000 a month. Two queries have already been created by the MIS department and are stored in the database: SalesbyMonth and SalesCountbyMonth. The first totals the dollar value; the second counts the number of transactions.
- Write a report to upper management designating the appropriate sales staff levels for each department by month. Include data and graphs to support your position. (Hint: Use a spreadsheet that lets you enter various staffing levels in each department in each month, and then calculate any sales declines.)
Technology Toolbox
- Create the PivotTable report for Rolling Thunder Bicycles. Briefly summarize any patterns or problems you identify.
- Using the Rolling Thunder Bicycles query, create a PivotChart and compare sales of the different models over time. Identify any patterns that you see.
- Research an alternate cube browser (such as SQL Server or Oracle) and explain how it is different from the Excel PivotTable. If you have access to the tool, build a small example.
- Compute the average number of days it takes to build a bicycle (ShipDate – OrderDate) for each month. Import the data into Excel and forecast the trend. First, forecast it based on all of the data. Second, forecast it for three time periods: (a) the early years, (b) the middle years, and (c) the most recent years. Look at the initial chart to estimate the breaks between these three sets, or just divide it into three equal-size groups if you do not see any good break points. Comment on any differences or problems.
- Using federal data (start at www.fedstats.gov), compute a regression analysis of Rolling Thunder sales by state by year compared with at least population and income.
Teamwork
- Have each person find and describe a problem that could benefit from a GIS. Make sure it needs a GIS, not just a mapping system. Combine the results and compare the types of problems to identify similarities.
- Work in pairs and have one person describe how to solve a problem that requires conditions while the second person creates the decision tree for the problem. For example, describe how to register for classes, or how to search for and purchase a product online.
- Each person should identify a problem that could be solved with a rules-based expert system. Combine the results and compare the types of problems to identify similarities.
- Using Rolling Thunder Bicycles, have each person forecast the sales by one model type for six months. Combine the individual model results and compare this value to the forecast based only on total sales.
- Select an economic data series such as personal income (check www.fedstats.gov). Place members into one of three subgroups. Have each group forecast the series using a different methodology. Compare the results. If you have sufficient data, leave out the most recent data, and then forecast those values and compare the forecasts to the actual.
- Choose a publicly-traded company and collect basic quarterly financial data for the company for at least 10 years. It is easiest if you assign specific years to each person. Put the data into a spreadsheet or simple database. Create a PivotTable and Pivot Chart to examine the data. Create a report containing some of the charts and tables and any conclusions you can make.
- Have each person find a problem that could benefit from a neural network. Describe how the system would be trained. Combine the individual comments and identify any commonalities.
Rolling Thunder Database
- Identify shipments where receipts do not match the original order. Provide a count and value (and percentages) by supplier/manufacturer.
- Analyze sales and discounts by employee and by model type. Are some employees providing higher discounts than others? Are we discounting some models too much or not enough?
- The company wants to create an online ordering system. Create a decision tree to help novices select the appropriate bicycle and components. If necessary, consult with a friend or relative who can be considered a bicycle expert.
- Use queries to extract sales data by model type and week. Use a spreadsheet to forecast the sales of each quantity of model type by week for the next year. Hint: Use Right("0" & Format([OrderDate],"ww"),2) to get the week number.
- What pattern-matching types of decisions arise at Rolling Thunder that could benefit from the use of neural networks?
- What aspects of customer service might be automated with expert systems? What are the potential advantages and disadvantages?