I. ordinal value.

Ii. discrete value.

Iii. continuous value.

Iv. nominal value.

I. values that are yielded from intended manipulation.

Ii. measurement from different target sources.

Iii. variability in measurement due to intentional intervention.

Iv. values of attributes that are deviated from other attributes.

I. multimodal distributions indicate there are multiple different patterns of population.

Ii. the probability of having x successful outcomes in n trials is defined as 𝑛𝐶𝑥 𝑝^𝑥(1-p)^(x–n).

Iii. almost all data falls within three standard deviations of the mean.

Iv. the distribution of the sample mean can be approximated by the normal distribution as the sample size gets "large enough."

（含图）

I. The null hypothesis, or the default assumption when flipping a coin several times, is that we are more likely to see either heads or tails.

II. A p-value greater than the significance level indicates that we can reject the null hypothesis.

III. A paired t-test should be used to compare two test scores of refrigerators made by different manufacturers under the same test conditions.

IV. The F-ratio will be large when there is a significant difference between groups due to a large variation between them.

Options:

.

I. Cause-effect relationship is essential in machine learning to explain why something happens.

II. Training data and test data are independent and identically distributed when the attributes of the data are related and the law that generates the data is the same.

III. Machine learning eliminates the differences in data, reduces the errors and finds the mapping between inputs and outputs.

IV. Independent and identically distributed is an important assumption to build, evaluate and generalize a predictive model.

I. linear regression makes more assumptions about the data resulting in low bias and high variance models.

II. non-linear machine learning algorithms often have more parameters to reduce bias errors.

III. low bias means there are fewer assumptions about the data and it can capture the complexity of the data.

IV. low variance means the prediction will not change much with changes in the data.

Brks <- c(0, 50, 80, 100)

Data$$grades <- cut(data$$scores, breaks=brks, include.lowest=T)

I. k can be terminated when residual sum of squares (RSS) falls above a threshold.

II. Higher value of k will result in higher confidence on the cross-validation result as compared to lower value of k.

III. k is calculated based on the distance between two nodes.

IV. k-fold validation reduces overfitting but does not completely eliminate it.

（含图）