Data Mining Assignment 2: Sample Input

Sample Files:

Test Files:

Sample Input 1:
We observe the work of customer-service representatives and identify situations that lead to customer complaints; thus, we view complaints as positive examples. We describe each observation by six attributes.

    Service area: rural or urban
    Temperature: cold, warm, or hot
    Weather: sunny, cloudy, or rainy
    Number of representatives: understaffed, normal, or overstaffed
    Day of the week: weekday or weekend
    Number of customers: small, medium, or large

The training examples show that we get complaints in urban areas, when an understaffed service has to face a large number of customers. The candidate-elimination algorithm converges to one hypothesis:

urban ? ? understaffed ? large

This hypothesis leads to the following classification of the test instances:

urban cold rainy understaffed weekday large positive

urban warm cloudy normal weekend large negative

Sample Input 2:
If a grocery store reduces the price of some items, it may increase the volume of sales. We need to identify the combinations of reduced-price products that increase sale; thus, we view an increase in the sales volume as a positive example. We describe a combination of products by five attributes.

    Meat: steak, chicken, or beef
    Dairy: cheese or milk
    Produce: carrots or peas
    Cereal: healthy or junk
    Snacks: chips, cookies, or candy

The candidate-elimination algorithm cannot find any hypothesis consistent with the training examples, and it terminates with a failure.

Sample Input 3:
We next consider six factors that may affect the profits of a department store, and we view profitable stores as positive examples.

    Neighborhood: poor, average, or rich
    Surrounding buildings: houses or apartments
    Age of residents: young, middle, or old
    Waterfront: nowater, lake, or beach
    Area: rural or urban
    Housing costs: low, medium, or high

The candidate-elimination algorithm finds several hypotheses consistent with the training examples.

Most specific hypothesis:

rich ? middle beach ? low

Most general hypotheses:

rich ? ? beach ? ?

? ? middle beach ? ?

? ? ? ? ? low

These hypotheses lead to the following classification of the test instances:

rich houses middle beach rural low positive

rich apartments middle nowater urban medium negative

average houses middle nowater rural low unknown

Back to the Data Mining home page
.

urban	cold	rainy	understaffed	weekday	large		positive
urban	warm	cloudy	normal	weekend	large		negative

rich	houses	middle	beach	rural	low	positive
rich	apartments	middle	nowater	urban	medium	negative
average	houses	middle	nowater	rural	low	unknown