免责声明：这是家庭作业的一部分。
我有一组推文，我需要创建一个分类器以尝试预测它们的情绪。我将通过创建“单词袋”模型并将径向SVM内核函数应用于数据来实现此目的。
以下是给您一个想法的原始数据：
> original_tweets
# A tibble: 2,385 x 3
tweet_id sentiment text
<int> <chr> <chr>
1 1 positive @TylerSkewes: It is almost 2014. Where are the self-driving cars so we don't have to worry about a DD tonight. Forreal tho
2 2 positive @WIRED: BMW builds a self-driving car -- that drifts I love this technology. Drive me to work baby!
3 3 positive Google better hurry up with that driverless car. Watching grandma do an 8 point turn to get in a parking spot is horrific.
4 4 positive I just waved thank you to this lady that let me merge on the highway and she gave me the finger. Need my self driving car.
5 5 positive I might be the only person who starts #cheering in their car when they see a @google car :) #happiness #feelslikeChristmas
6 6 positive I want the driverless car, and BAD. Seriously I would be happy if tomorrow morning there were no drivers behind the wheel.
7 7 positive I'm over here writing a 2000 word essay while *****s at Google are on driverless cars making ground breaking shit. Damn. _
8 8 positive Is it crazy to think that self driving cars will be the biggest innovation of the last few decades?
9 9 positive Its very nice!RT @cdixon: It's awesome that Google is investing in futuristic stuff like AR glasses and self-driving cars.
10 10 positive Look closely you will see the reflection of a google car !!!! Screen shot from google maps !!!!!
# ... with 2,375 more rows
>
我稍微修改了一些术语，因为它们中包含URL，但是您明白了。
我已经将数据格式化为整齐的格式，并计算了每个词的TF-IDF分数。对于我的功能空间，我采用了IDF得分最高的1000个词。
以下是我的数据示例：
> feature_space
# A tibble: 3,000 x 7
tweet_id sentiment word n tf idf tf_idf
<int> <chr> <chr> <int> <dbl> <dbl> <dbl>
1 1 positive forreal 1 0.0435 7.78 0.338
2 2 positive drifts 1 0.0476 7.78 0.370
3 2 positive rprjtelkg6 1 0.0476 7.78 0.370
4 5 positive cheering 1 0.0455 7.78 0.353
5 5 positive feelslikechristmas 1 0.0455 7.78 0.353
6 7 positive 2000 1 0.0476 7.78 0.370
7 7 positive *****s 1 0.0476 7.78 0.370
8 8 positive decades 1 0.0417 7.78 0.324
9 8 positive vltlymug89 1 0.0417 7.78 0.324
10 9 positive ar 1 0.0476 7.78 0.370
# ... with 2,990 more rows
我想使用他们的TF-IDF分数创建一个词袋模型来创建情感分类器。对于这种模型，我知道我需要设置数据框架，以使特征空间中每个可能的TF-IDF术语权重的每一条鸣叫都是一行，一列。
我很难弄清楚如何最好地改变小节或数据帧以使数据变为这种格式。我已经尝试过mutate（）和join（）的各种组合，但这并不是我想要的。
如何基于一组功能词快速将3000列或更多列添加到数据框或小标题，并应用其TF-IDF值来填充此稀疏数据结构？我不一定需要直接的代码答案，但是朝着正确的方向迈出如何在R中实现这一点的步骤将对我有很大帮助。
更新：我现在的单词袋里有一个空的小标题，我只需要在数据中填写非零的TF-DF值即可。在这里：
> bag_of_words
# A tibble: 2,385 x 3,002
tweet_id sentiment forreal drifts rprjtelkg6 cheering feelslikechristmas `2000` *****s decades vltlymug89 ar closely reflection zg7hvvfgpn
<int> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 1 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
2 2 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
3 3 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
4 4 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
5 5 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
6 6 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
7 7 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
8 8 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
9 9 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
10 10 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
# ... with 2,375 more rows, and 2,987 more variables
答案 0 :(得分：0)
好的，我想我有一个解决方案。我绝对很好奇如何在没有for循环的情况下实现此功能，但我仍然对apply（）样式的编码不太满意。
这是我想出的：
#create bag of words model
#get tweet_id and sentiment
bag_of_words <- original_tweets %>%
select(-one_of('text'))
#get words from feature space
feature_words <- feature_space$word
#generate empty columns
for(i in feature_words)
bag_of_words[,i] <- 0
#fill in columns with values from feature space
for(i in 1:length(feature_words)) {
word <- feature_space[i,]$word
tweet <- feature_space[i,]$tweet_id
score <- feature_space[i,]$tf_idf
bag_of_words[tweet,word] <- score
}
检查输出，看起来不错：
> bag_of_words
# A tibble: 2,385 x 3,002
tweet_id sentiment forreal drifts rprjtelkg6 cheering feelslikechristmas `2000` *****s decades vltlymug89 ar closely reflection zg7hvvfgpn
<int> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 1 positive 0.338 0 0 0 0 0 0 0 0 0 0 0 0
2 2 positive 0 0.370 0.370 0 0 0 0 0 0 0 0 0 0
3 3 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
4 4 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
5 5 positive 0 0 0 0.353 0.353 0 0 0 0 0 0 0 0
6 6 positive 0 0 0 0 0 0 0 0 0 0 0 0 0
7 7 positive 0 0 0 0 0 0.370 0.370 0 0 0 0 0 0
8 8 positive 0 0 0 0 0 0 0 0.324 0.324 0 0 0 0
9 9 positive 0 0 0 0 0 0 0 0 0 0.370 0 0 0
10 10 positive 0 0 0 0 0 0 0 0 0 0 0.370 0.370 0.370
# ... with 2,375 more rows, and 2,987 more variables
回想起来，我对自己的要求可能会比我需要做的要难，但是我绝对希望看到可以采用任何更有效的方法来完成此经验丰富的R vets。干杯。