特定列的R apriori规则

Question

I have a following dataset:

from | to | time_period
house | shop | evening
residential building | transportation | night
....
food | public building | morning

I use Apriori algorithm:

rules = apriori(data, parameter=list(support=0.01, confidence=0.5));
inspect(head(sort(rules, by="lift"),10));

And it produces me following output: lhs rhs support confidence lift

1  {from=residential building,                                                              
    to=food}                   => {time_period=night}         0.01398601  0.5882353 2.285806
2  {from=entertainment}        => {time_period=evening}       0.02517483  0.5294118 2.188031
3  {to=entertainment}          => {time_period=evening}       0.01678322  0.5217391 2.156321
4  {to=food,                                                                                
    time_period=night}         => {from=residential building} 0.01398601  1.0000000 1.735437
5  {from=food,                                                                              
    to=food}                   => {time_period=daytime}       0.01538462  0.8461538 1.689944
6  {from=public building,                                                                   
    to=food}                   => {time_period=daytime}       0.01958042  0.8235294 1.644758
7  {to=residential building,                                                                
    time_period=night}         => {from=residential building} 0.19580420  0.9459459 1.641629
8  {time_period=night}         => {from=residential building} 0.24195804  0.9402174 1.631688
9  {from=education,                                                                         
    to=residential building}   => {time_period=daytime}       0.01538462  0.7857143 1.569234
10 {from=food,                                                                              
    to=residential building}   => {time_period=daytime}       0.02237762  0.7619048 1.521681

It generates all kinds of rules, but this is not exactly I want. I want to have only rules of

{from, time_period} => {to}

For example, {from=food, time_period=daytime} => {to=residential building}. I am not interested in any other rules except {from, time_period} => {to} , which means that rules like {from, to} => {time_period} or {time_period} => {from} or any other does not interest me.

How can I do that?

Answer 1

Just filter the results , and keep only those rules that you are interested in.

The expensive part of Apriori is finding the frequent itemsets, and you cannot save much there. You do also need the frequency of itemsets without to .

Generating the actual rules afterwards is cheap; so you might just as well generate all of them, then only keep those with a to on the right.

However, given that you only have three columns, and one is your desired outcome, you don't need association rule mining at all .

Apriori and these are beneficial when you have long rules, because they can avoid redundant computations. Apriori starts getting interesting at length 3 the earliest.

Here, Apriori will not be more efficient than just enumerating all from -> to , time_period -> to and from + time_period -> to combinations and computing the desired quality each.

Answer 2

I came across the same situation where i needed to find relationship with only two attributes where the data frame had 11 attributes in total. Using apriori generated roughly 7000 rules.

Try something like below. It worked for me.

rules = apriori( select(data,c("from","time_period","to") ), parameter=list(support=0.01, confidence=0.5));

Input for the apriori function is only the columns of interest. I extracted columns using select in dplyr package.