欺詐行為識(shí)別

背景 (Background)

Online recruitment fraud (ORF) is a form of malicious behaviour that aims to inflict loss of privacy, economic damage or harm the reputation of the stakeholders via fraudulent job advertisements.

在線招聘欺詐(ORF)是一種惡意行為，旨在通過欺詐性的招聘廣告造成隱私的喪失，經(jīng)濟(jì)損失或損害利益相關(guān)者的聲譽(yù)。

The aim of the analytics task was to identify fraudulent job advertisements from the data, determine key indicators of fraud and make recommendations on how to identify fraudulent job advertisements in the future.

分析任務(wù)的目的是從數(shù)據(jù)中識(shí)別欺詐性工作廣告，確定欺詐的關(guān)鍵指標(biāo)，并就未來如何識(shí)別欺詐性工作廣告提出建議。

數(shù)據(jù)集 (Dataset)

We will use the Employment Scam Aegean Dataset (EMSCAD), which can be downloaded at http://icsdweb.aegean.gr/emscad. A description of how the data was collected and a data dictionary is available on this page.

我們將使用“就業(yè)騙局愛琴海數(shù)據(jù)集(EMSCAD)”，該數(shù)據(jù)集可從http://icsdweb.aegean.gr/emscad下載。 此頁面上提供了有關(guān)如何收集數(shù)據(jù)和數(shù)據(jù)字典的描述。

The dataset contains 17,880 real-life job ads. Variables within the dataset include:

數(shù)據(jù)集包含17,880個(gè)現(xiàn)實(shí)工作廣告。 數(shù)據(jù)集中的變量包括：

方法 (Methodology)

It is firstly important to understand how the dataset can be utilized to distinguish between fraudulent and non-fraudulent ads as this will govern the type of analytical method will be employed. The response variable which is the binary field “fraudulent” is what we are trying to predict where t = “Yes” and f = “No”.

首先，重要的是要了解如何利用數(shù)據(jù)集來區(qū)分欺詐性廣告和非欺詐性廣告，因?yàn)檫@將決定采用何種分析方法。 我們?cè)噲D預(yù)測的響應(yīng)變量是二進(jìn)制字段“欺詐性”，其中t =“是”而f =“否”。

了解數(shù)據(jù)集以選擇分析方法 (Understanding the dataset to choose the analytical approach)

We have a set of variables which are not categorized and are essentially HTML strings — Benefits, Company profile, Description and Requirements. The type of analysis that will be required of textual data is Sentiment & Emotion analysis or Frequency analysis.

我們有一組未分類的變量，它們實(shí)質(zhì)上是HTML字符串-收益，公司簡介，描述和要求。 文本數(shù)據(jù)所需的分析類型為情感與情感分析或頻率分析。

The categorical or factor variables of which there are 11 — location, company logo, industry, function, salary range, department, required education, required experience, employment type, telecommuting, and questions will be inputs into a machine learning algorithm such as the gradient boosting machine (GBM), distributed random forest (DRF) and generalized linear model (GLMNET) to determine the top predictors that can be used to distinguish between fraudulent and non-fraudulent ads.

其中包含11個(gè)類別或類別變量-位置，公司徽標(biāo)，行業(yè)，職能，薪資范圍，部門，所需教育，所需經(jīng)驗(yàn)，就業(yè)類型，遠(yuǎn)程辦公和問題將輸入到機(jī)器學(xué)習(xí)算法中，例如漸變提升機(jī)(GBM)，分布式隨機(jī)森林(DRF)和廣義線性模型(GLMNET)來確定可用于區(qū)分欺詐性廣告和非欺詐性廣告的最佳預(yù)測指標(biāo)。

As two analytical approaches will be used — one for the string variables and the other for the factor variables, there will be two sets of output as follows:

由于將使用兩種分析方法(一種用于字符串變量，另一種用于因子變量)，因此將有兩組輸出，如下所示：

• HTML variables: Sentiment, emotion and word frequency plots
  HTML變量：情感，情感和單詞頻率圖
• Nominal & Binary Variables: Top predictors, coefficients
  標(biāo)稱和二進(jìn)制變量：最佳預(yù)測變量，系數(shù)

Some variables do not contribute to any information and as such, they were excluded from the analysis. These include title and in-balanced as title is identifying information and in-balanced is used to include and exclude records to balance the dataset.

一些變量不會(huì)提供任何信息，因此將其從分析中排除。 這些包括標(biāo)題和不平衡，因?yàn)闃?biāo)題是標(biāo)識(shí)信息，不平衡用于包括和排除記錄以平衡數(shù)據(jù)集。

數(shù)據(jù)ETL (Data ETL)

Prior to modelling, a number of steps have to be carried out to cleanse the dataset. The flow diagram below shows the steps that were carried out to prepare the dataset for modelling.

在建模之前，必須執(zhí)行許多步驟來清理數(shù)據(jù)集。 下面的流程圖顯示了為準(zhǔn)備建模數(shù)據(jù)集而執(zhí)行的步驟。

Figure 1: Data ETL Flow Diagram圖1：數(shù)據(jù)ETL流程圖

結(jié)果 (Results)

輸出-文本分析 (Output — Text Analytics)

詞云 (Word clouds)

Word clouds were created for each of the HTML strings — company profile, job description, requirements and benefits.

為每個(gè)HTML字符串創(chuàng)建了文字云-公司簡介，職位描述，要求和收益。

The word clouds below are for company profile for non-fraudulent ads (left) and fraudulent ads (right).

下面的詞云是針對(duì)非欺詐性廣告(左)和欺詐性廣告(右)的公司資料。

• Non-fraudulent ads emphasize a work life balance (“home”, “l(fā)ife”, “care”) and company culture (“team”, “experience”)
  不欺詐的廣告強(qiáng)調(diào)工作與生活之間的平衡(“家庭”，“生活”，“護(hù)理”)和公司文化(“團(tuán)隊(duì)”，“體驗(yàn)”)
• Fraudulent ads are largely missing the company profile with an emphasis on monetary perks (“cell phones”, “money”, “cost”)
  欺詐性廣告在很大程度上缺少公司形象，而側(cè)重于金錢利益(“手機(jī)”，“金錢”，“成本”)

Figure 2.1: Company Ads — Non-Fraudulent圖2.1：公司廣告-非欺詐性 Figure 2.2: Company Ads — Fraudulent圖2.2：公司廣告-欺詐

The word clouds below are for job description for non-fraudulent ads (left) and fraudulent ads (right).

以下單詞云用于描述非欺詐性廣告(左)和欺詐性廣告(右)的工作。

• Non-fraudulent ads emphasize company offerings (“gas”, “oil”, “operations”)
  非欺詐性廣告強(qiáng)調(diào)公司產(chǎn)品(“天然氣”，“石油”，“運(yùn)營”)
• Fraudulent ads emphasize monetary value (“money”, “financially”, “discounts”) Non-fraudulent ads emphasize company offerings (“gas”, “oil”, “operations”)
  欺詐性廣告強(qiáng)調(diào)貨幣價(jià)值(“金錢”，“財(cái)務(wù)”，“折扣”)欺詐性廣告強(qiáng)調(diào)公司產(chǎn)品(“天然氣”，“石油”，“運(yùn)營”)

Figure 3.1: Job description — Non-fraudulent ads圖3.1：職位描述-非欺詐性廣告 Figure 3.2: Job description: Fraudulent Ads圖3.2：職位描述：欺詐性廣告

Below are word clouds for job requirements — non-fraudulent on the left and fraudulent on the right.

以下是工作要求的詞云-左側(cè)為欺詐性質(zhì)，右側(cè)為欺詐性質(zhì)。

• Non-fraudulent ads emphasize years of experience, skills, degree qualifications and project orientation
  非欺詐性廣告強(qiáng)調(diào)多年的經(jīng)驗(yàn)，技能，學(xué)位資格和項(xiàng)目方向
• Fraudulent ads emphasize the above attributes to a lesser extent Non-fraudulent ads emphasize years of experience, skills, degree qualifications and project orientation
  欺詐性廣告較少強(qiáng)調(diào)上述屬性欺詐性廣告強(qiáng)調(diào)較少的經(jīng)驗(yàn)，技能，學(xué)位資格和項(xiàng)目方向

Figure 4.1: Job requirements — Non-fraudulent ads圖4.1：職位要求-不欺詐的廣告 Figure 4.2: Job requirements — Fraudulent ads圖4.2：職位要求-欺詐性廣告

Finally, the word clouds below are based on the test for job benefits.

最后，下面的“云”一詞基于對(duì)工作福利的測試。

• Non-fraudulent ads emphasize benefits such as “sick leave”, “hours” and “vacation”
  不欺詐的廣告會(huì)強(qiáng)調(diào)諸如“請(qǐng)病假”，“工作時(shí)間”和“假期”之類的好處
• Fraudulent ads appear to offer monetary perks such as accommodation, holidays, food, competitive salary, visa, and food among others.
  欺詐性廣告似乎提供金錢福利，例如住宿，假期，食物，有競爭力的薪水，簽證和食物等。

Figure 5.1: Job benefits — Non-fraudulent ads圖5.1：工作收益-非欺詐性廣告 Figure 5.2: Job benefits — Fraudulent ads圖5.2：工作收益-欺詐性廣告

情緒分析 (Sentiment Analysis)

Another way to analyse text is via sentiment analysis, which is type of emotion (positive or negative) associated with each word in text.

分析文本的另一種方法是通過情感分析，這是與文本中每個(gè)單詞相關(guān)的情感類型(正面或負(fù)面)。

For instance, looking at the emotion categories below for non-fraudulent and fraudulent ads for job requirements, we can see that a greater proportion of non-fraudulent ads (left) are positive (“joy”, “surprise”), whereas the contrast is true for fraudulent ads (right).

例如，查看下面針對(duì)工作要求的不欺詐和欺詐廣告的情感類別，我們可以看到，較大比例的不欺詐廣告(左)是積極的(“歡樂”，“驚奇”)，而對(duì)比對(duì)于欺詐性廣告是正確的(右)。

Figure 6: Emotion sentiment analysis: Non-fraudulent vs. Fraudulent ads圖6：情緒情感分析：非欺詐性廣告與欺詐性廣告

We can also look at the polarity of these ads that is the orientation towards a specific emotion category, positive or negative. A greater proportion of non-fraudulent ads are positive than fraudulent ads.

我們還可以查看這些廣告的極性，即針對(duì)特定情緒類別(正面或負(fù)面)的方向。 與欺詐性廣告相比，非欺詐性廣告中肯定的比例更大。

Figure 7: Emotion (Polarity) analysis for non-fraudulent vs. fraudulent ads圖7：非欺詐性廣告與欺詐性廣告的情感(極性)分析

As shown from the examples of word clouds and bar graphs of textual sentiment, we can see that text information is very useful in predicting certain behaviour. The next logical step would be to tag these ads as positive or negative based on their emotion/ polarity and introduce this information as binary variables into a machine learning model for prediction to determine the importance of these variables for prediction.

從單詞云和文本情感條形圖的示例中可以看出，文本信息對(duì)于預(yù)測某些行為非常有用。 下一步的邏輯步驟是根據(jù)廣告的情感/極性將這些廣告標(biāo)記為肯定或否定，并將此信息作為二進(jìn)制變量引入到機(jī)器學(xué)習(xí)模型中進(jìn)行預(yù)測，以確定這些變量對(duì)預(yù)測的重要性。

For instance, you would create four variables, job requirements, description, benefits and company profile. For each variable, each ad would be assigned a “0” or “1” to signify “positive” or “negative” sentiment.

例如，您將創(chuàng)建四個(gè)變量，工作要求，描述，福利和公司簡介。 對(duì)于每個(gè)變量，將為每個(gè)廣告分配“ 0”或“ 1”，以表示“積極”或“消極”情緒。

Now, let’s move on to utilizing the numerical variables in a model for predicting which ads are fraudulent and non-fraudulent.

現(xiàn)在，讓我們繼續(xù)使用模型中的數(shù)字變量來預(yù)測哪些廣告是欺詐性和非欺詐性的。

機(jī)器學(xué)習(xí)模型 (Machine Learning Models)

總覽 (Overview)

It is always good to run a number of different types of models and then select the one or combination of models that provide you not only with the highest accuracy but also meaningful results that can readily be explained to business stakeholders and are likely to be accepted by them.

最好運(yùn)行多個(gè)不同類型的模型，然后選擇一個(gè)或多個(gè)模型組合，這些模型不僅可以為您提供最高的準(zhǔn)確性，而且還可以向業(yè)務(wù)利益相關(guān)方解釋并可以為您所接受的有意義的結(jié)果。他們。

For this problem, I ran three types of models:

對(duì)于這個(gè)問題，我運(yùn)行了三種類型的模型：

• Distributed random forest (DRF): Essentially a random forest which is an ensemble of classification trees but run in parallel on the h2o server, hence, the word distributed.

  分布式隨機(jī)森林 ( DRF )：本質(zhì)上是一個(gè)隨機(jī)森林，它是分類樹的集合，但是在h2o服務(wù)器上并行運(yùn)行，因此是分布式的。

• Gradient boosting machine (GBM): Like the random forest, it is also a classification method consisting of an ensemble of trees. The difference is that random forests are used to build deep independent trees (i.e. each tree is run on a random set of variables on a random subset of the data — the “bagging” method), whereas GBMs built lots of shallow and weak, dependent, successive trees. In this approach, each tree learns from the previous tree and tries to improve on it by reducing the amount of error and increasing the amount of variation in the response variable explained by the predictive variables.

  梯度提升機(jī) ( GBM )：與隨機(jī)森林一樣，它也是一種由樹木集合組成的分類方法。 不同之處在于，隨機(jī)森林用于構(gòu)建深層獨(dú)立的樹(即，每棵樹都在數(shù)據(jù)的隨機(jī)子集上的隨機(jī)變量集上運(yùn)行-“裝袋”方法)，而GBM則構(gòu)建了許多淺層和弱層，相關(guān)的，連續(xù)的樹木。 在這種方法中，每棵樹都從前一棵樹中學(xué)習(xí)，并嘗試通過減少錯(cuò)誤量和增加由預(yù)測變量解釋的響應(yīng)變量的變化量來對(duì)其進(jìn)行改進(jìn)。

• Generalized linear model (GLM): GLMs are just an extension of linear models that can be run on a non-normally distributed dependent variable. As this is a classification problem, the link function used is for logistic regression. The output of a logistic regression algorithm are coefficients for the predictor in logits, where a one unit change in the predictor variable leads to the coefficient value change in the log odds. These logits can be converted to odds ratio to provide more meaningful information.

  廣義線性模型 ( GLM )： GLM只是線性模型的擴(kuò)展，可以在非正態(tài)分布的因變量上運(yùn)行。 由于這是一個(gè)分類問題，因此使用的鏈接函數(shù)用于邏輯回歸 。 Logistic回歸算法的輸出是logits中預(yù)測變量的系數(shù)，其中預(yù)測變量的單位變化導(dǎo)致對(duì)數(shù)賠率的系數(shù)值變化。 可以將這些logit轉(zhuǎn)換為優(yōu)勢比，以提供更多有意義的信息。

• To calculate the odds ratio, we need to exponentiate each coefficient by raising it to the power of e i.e. e^b

  要計(jì)算比值比，我們需要通過將每個(gè)系數(shù)提高到e的冪( 即e ^ b)來取冪

Now that you have some understanding of the three types of models, let’s compare their model accuracy.

現(xiàn)在您已經(jīng)對(duì)這三種類型的模型有了一定的了解，讓我們比較它們的模型準(zhǔn)確性。

方法 (Methodology)

The dataset was split into training (80% of dataset) and test (20% of dataset) sets using a random seed where the goal is to train the model on the training set and test its accuracy on the test set.

使用隨機(jī)種子將數(shù)據(jù)集分為訓(xùn)練集(占數(shù)據(jù)集的80％)和測試集(占數(shù)據(jù)集的20％)，其目的是在訓(xùn)練集上訓(xùn)練模型并在測試集上測試其準(zhǔn)確性。

The GBM was run with the following parameters where the max depth of the tree was set to 4 (4 levels), a small learn rate, and five fold cross validation.

使用以下參數(shù)運(yùn)行GBM，其中樹的最大深度設(shè)置為4(4個(gè)級(jí)別)，學(xué)習(xí)率小，交叉驗(yàn)證五倍。

Cross-validation is a technique used to validate our training model before we apply it to the test set. By specifying five folds, it means that we build five different models where each model is trained on four parts and tested on the fifth. So, the first model is trained on parts 1, 2, 3, and 4 and tested on 5. The second model is trained on parts 1, 3, 4, and 5 and tested on part 2 and so on.

交叉驗(yàn)證是一種用于將訓(xùn)練模型應(yīng)用于測試集之前對(duì)其進(jìn)行驗(yàn)證的技術(shù)。 通過指定五折，這意味著我們建立了五個(gè)不同的模型，其中每個(gè)模型分為四個(gè)部分進(jìn)行訓(xùn)練，并在第五個(gè)部分進(jìn)行測試。 因此，第一個(gè)模型在零件1、2、3和4上進(jìn)行訓(xùn)練，并在5上進(jìn)行測試。第二個(gè)模型在零件1、3、4和5上進(jìn)行訓(xùn)練，并在第2部分上進(jìn)行測試，依此類推。

This method is called k-fold cross-validation and allows us to be more confident in the performance of the modelling method utilised. When we create five different models, we are testing it on five different/unseen datasets. If we only test the model once, for example, on our test set, then we only have a single evaluation which may be a biased results.

這種方法稱為k折交叉驗(yàn)證，它使我們對(duì)所使用的建模方法的性能更有信心。 當(dāng)我們創(chuàng)建五個(gè)不同的模型時(shí)，我們正在五個(gè)不同/看不見的數(shù)據(jù)集上對(duì)其進(jìn)行測試。 例如，如果僅在測試集上對(duì)模型進(jìn)行一次測試，則只有一個(gè)評(píng)估，這可能是有偏差的結(jié)果。

gbm_model <-h2o.gbm(y=y_dv, x=x_iv, training_frame = model_train.h2o,
ntrees =500, max_depth = 4, distribution="bernoulli", #for 0-1 outcomes
learn_rate = 0.01, seed = 1234, nfolds = 5, keep_cross_validation_predictions = TRUE)

To measure model accuracy, I used the ROC-AUC metrics. ROC or Receiver Operating Characteristic is a probability curve and the AUC, Area Under Curve, is a measure of the degree of separation between classes. In our case, the AUC is how accurately can the given model distinguish between non-fraudulent and fraudulent ads. The higher the AUC, the more accurate the model is at classifying the ads correctly.

為了測量模型的準(zhǔn)確性，我使用了ROC-AUC指標(biāo)。 ROC或接收器工作特性是一條概率曲線，而AUC(曲線下面積)是對(duì)類別之間分離程度的度量。 在我們的案例中，AUC是給定模型在非欺詐性廣告和欺詐性廣告之間的區(qū)分精度。 AUC越高，模型正確分類廣告的準(zhǔn)確性就越高。

fpr <- h2o.fpr( h2o.performance(gbm_model, newdata=model_test.h2o) )[['fpr']]
tpr <- h2o.tpr( h2o.performance(gbm_model, newdata=model_test.h2o) )[['tpr']]
ggplot( data.table(fpr = fpr, tpr = tpr), aes(fpr, tpr) ) +
geom_line() + theme_bw() + ggtitle( sprintf('AUC: %f', gbm.auc) )

AUC is made up of a couple of metrics to test model accuracy which are:

AUC由幾個(gè)衡量模型準(zhǔn)確性的指標(biāo)組成：

• True Positives (TP): Fraudulent ads that were correctly predicted as fraudulent
  真實(shí)肯定(TP)：被正確預(yù)測為欺詐的欺詐性廣告
• True Negatives (TN): Non-fraudulent ads that were correctly predicted as non-fraudulent
  真實(shí)否定詞(TN)：正確預(yù)測為非欺詐性的非欺詐性廣告
• False Positives (FP): Non-fraudulent ads that were incorrectly predicted as fraudulent
  誤報(bào)(FP)：被誤認(rèn)為是欺詐的非欺詐性廣告
• False Negatives (FN): Fraudulent ads that were incorrectly predicted as non-fraudulent
  假陰性(FN)：被錯(cuò)誤地預(yù)測為非欺詐的欺詐廣告

These metrics can then be combined to calculate sensitivity and specificity.

然后可以將這些指標(biāo)進(jìn)行組合以計(jì)算敏感性和特異性。

Sensitivity is a measure of what proportion of fraudulent ads were correctly classified.

敏感性衡量正確分類欺詐廣告的比例。

Sensitivity = count (TP) / sum(count(TP) + count(FP))

靈敏度=計(jì)數(shù)(TP)/總和(計(jì)數(shù)(TP)+計(jì)數(shù)(FP))

Specificity is a measure of what proportion of non-fraudulent ads were correctly identified.

特異性是衡量正確識(shí)別非欺詐性廣告比例的一種方法。

Specificity = count (FP)/sum (count(TP) + count(FP))

特異性=計(jì)數(shù)(FP)/總和(計(jì)數(shù)(TP)+計(jì)數(shù)(FP))

When determining which measure is more important for your analysis, ask yourself the question whether it is more important for you to identify the number of correctly classified positives (sensitivity is more important) or negatives (specificity is more important). In our case, we want a model with higher sensitivity as we are more interested in correctly distinguishing fraudulent ads.

在確定哪種量度對(duì)您的分析更重要時(shí)，問自己一個(gè)問題，即確定正確分類的陽性(敏感性更重要)或陰性(特異性更重要)的數(shù)量對(duì)您來說更重要。 在我們的案例中，我們希望模型具有更高的靈敏度，因?yàn)槲覀儗?duì)正確區(qū)分欺詐性廣告更加感興趣。

All these metrics can be summarized in a confusion matrix which is a table comparing number of cases that were correctly and incorrectly predicted against the actual number of fraudulent and non-fraudulent cases. This information can be used to supplement our understanding of the ROC and AUC metrics.

所有這些指標(biāo)都可以匯總在一個(gè)混淆矩陣中 ，該矩陣是一個(gè)表格，該表格將正確和錯(cuò)誤地預(yù)測的案件數(shù)量與欺詐和非欺詐案件的實(shí)際數(shù)量進(jìn)行比較。 此信息可用于補(bǔ)充我們對(duì)ROC和AUC指標(biāo)的理解。

Another aspect of the ROC-AUC metrics is the threshold used to determine whether an ad is fraudulent or non-fraudulent. To determine the best threshold t that maximizes the number of TPs positives, we can use the ROC curve, where we plot the TPR (True Positive Rate) on the y-axis against the FPR (False Positive Rate) on the x-axis.

ROC-AUC指標(biāo)的另一個(gè)方面是用于確定廣告是欺詐還是不欺詐的閾值 。 為了確定使TP陽性數(shù)最大化的最佳閾值t，我們可以使用ROC曲線，在該曲線上我們繪制y軸上的TPR(真陽性率)相對(duì)于x軸上的FPR(假陽性率)。

The AUC allows for comparison of models where we can compare their ROC curves for model accuracy on the test set as shown in the model output below.

AUC允許對(duì)模型進(jìn)行比較，我們可以在測試集上比較其ROC曲線以確保模型準(zhǔn)確性，如下面的模型輸出所示。

模型輸出 (Model Output)

模型精度比較 (Model Accuracy Comparison)

The table below shows that the DRF produces a model with the highest AUC of 0.962 on the test set. All three models have high AUC values (> 0.5 or random prediction).

下表顯示了DRF在測試集上生成的AUC最高為0.962的模型。 這三個(gè)模型均具有較高的AUC值(> 0.5或隨機(jī)預(yù)測)。

Figure 9: AUC curve for classification of ads by fraudulence圖9：按欺詐分類廣告的AUC曲線

However, let’s dig deeper into what this AUC means in terms of correctly classified ads as fraudulent by looking at the confusion matrix below for the GLM model as an example.

但是，讓我們通過以GLM模型為例，查看下面的混淆矩陣，進(jìn)一步深入了解該AUC在將廣告正確分類為欺詐廣告方面的含義。

The confusion matrix for GLM on the test set indicates an error rate of 8.15% in classifying fraudulent cases incorrectly. The sensitivity for this model is 327/(327+29) = 92% which is very good.

測試集上的GLM混淆矩陣表明，錯(cuò)誤地對(duì)欺詐案件進(jìn)行分類的錯(cuò)誤率為8.15％。 該模型的靈敏度為327 /(327 + 29)= 92％，非常好。

Now, let’s look at the remaining output from the models, more specifically what are the top predictors in classifying fraudulent and non-fraudulent ads.

現(xiàn)在，讓我們看一下模型的其余輸出，更具體地說，是對(duì)欺詐性和非欺詐性廣告進(jìn)行分類的最佳預(yù)測指標(biāo)是什么。

最重要的預(yù)測因子 (Most Important Predictors)

The variable importance rank in a classification problem tells us how accurately can a predictor variable classify fraudulent ads over non-fraudulent ads relative to all other predictors that were used int he mode.

分類問題中變量的重要性等級(jí)告訴我們，與在模式中使用的所有其他預(yù)測變量相比，預(yù)測變量可以如何準(zhǔn)確地將欺詐性廣告分類為非欺詐性廣告。

For both (a) GBM and (b) DRF, the top three variables — location, company logo and industry — in terms of how useful they are in classifying job ads into fraudulent or non-fraudulent are the same. This is also true for the has questions and telecommuting variables as being least important

對(duì)于(a)GBM和(b)DRF，就它們?cè)趯⒄衅笍V告分類為欺詐或不欺詐方面的有用程度而言，前三個(gè)變量(位置，公司徽標(biāo)和行業(yè))是相同的。 對(duì)于具有最不重要的問題和遠(yuǎn)程辦公變量也是如此

Now, let’s plot the dataset to better understand how the top predictors vary for fraudulent and non-fraudulent ads.

現(xiàn)在，讓我們繪制數(shù)據(jù)集，以更好地了解欺詐性和非欺詐性廣告的主要預(yù)測變量如何變化。

Let’s look at the top variable, location, where we can see that a greater proportion of fraudulent than non-fraudulent ads are from the USA and Australia as indicated by the circled bars.

讓我們看一下最上面的變量location ，在該變量中，如帶圓圈的條所示，我們發(fā)現(xiàn)來自美國和澳大利亞的欺詐廣告比非欺詐廣告更大。

Figure 10: Frequency of Fraudulent vs. non-fraudulent ads by country圖10：按國家劃分的欺詐性廣告和非欺詐性廣告的頻率

A greater proportion of fraudulent than non-fraudulent ads do not display a company logo in their job ads.

欺詐性廣告中比不欺詐性廣告更大的比例在其招聘廣告中不顯示公司徽標(biāo)。

Figure 11: Frequency of fraudulent vs. non-fraudulent ads by absence/presence of company logo in ad圖11：廣告中是否存在公司徽標(biāo)的欺詐性廣告與非欺詐性廣告的頻率

了解模型系數(shù) (Understanding the model coefficients)

Now, let’s try to numerically understand the relationship between the predictor variables and the classification of ads.

現(xiàn)在，讓我們嘗試從數(shù)字上了解預(yù)測變量與廣告分類之間的關(guān)系。

As shown in the table below, the highlighted predictors are best at distinguishing fraudulent and non-fraudulent ads.

如下表所示，突出顯示的預(yù)測變量最能區(qū)分欺詐性和非欺詐性廣告。

• 767 variables entered into model, only 48 have a non-zero coefficient (Top predictors shown)

  輸入模型的767個(gè)變量中，只有48個(gè)具有非零系數(shù)(顯示了最高預(yù)測變量)

• The greater the probability, the higher the chance of the ad being fraudulent

  可能性越大，則廣告被欺詐的可能性越高。

結(jié)論和后續(xù)步驟 (Conclusion and Next Steps)

Now that you have a good understanding of using both textual and numerical predictors in a classification problem with the employment of both textual analytics tools and machine learning classification algorithms.

現(xiàn)在，您已經(jīng)對(duì)使用文本分析工具和機(jī)器學(xué)習(xí)分類算法同時(shí)使用文本預(yù)測器和數(shù)字預(yù)測器在分類問題中有了很好的了解。

So, what can we do next?

那么，下一步我們?cè)撛趺醋?#xff1f;

• A combination of textual analysis and predictive modelling should be used to classify job ads into fraudulent and non-fraudulent job ads
  應(yīng)結(jié)合使用文本分析和預(yù)測模型來將求職廣告分為欺詐性和非欺詐性職業(yè)廣告
• To improve accuracy of textual analysis the following methods can be introduced:
  為了提高文本分析的準(zhǔn)確性，可以引入以下方法：

• N-grams modelling: Look at the combination of words that occur together to identify patterns

  N-gram建模 ：查看一起出現(xiàn)的單詞組合以識(shí)別模式

• Look for trends in capitalization and punctuation
  尋找大寫和標(biāo)點(diǎn)符號(hào)的趨勢
• Look for trends in emphasized text (bold, italicized)
  在強(qiáng)調(diào)的文本中查找趨勢(粗體，斜體)
• Look for trends in types of HTML tags used (raw text lists vs. list text wrapped in list elements)
  尋找使用HTML標(biāo)簽類型的趨勢(原始文本列表與列表元素中包裹的列表文本)

Predictive model accuracy can be improved by:

預(yù)測模型的準(zhǔn)確性可以通過以下方法提高：

• Working with a larger dataset
  處理更大的數(shù)據(jù)集
• Splitting dataset into three: training, test and validation sets
  將數(shù)據(jù)集分為三部分：訓(xùn)練集，測試集和驗(yàn)證集
• Splitting salary range into numerical variables: minimum & maximum
  將薪水范圍分為數(shù)字變量：最小和最大
• Removing variables that are associated with each other (i.e. use of chi-squared test of independence)
  刪除相互關(guān)聯(lián)的變量(即，使用卡方獨(dú)立性檢驗(yàn))
• Splitting location into country, state, and city
  將位置分為國家，州和城市
• Reducing number of variables by grouping industry and function categories
  通過對(duì)行業(yè)和職能類別進(jìn)行分組來減少變量數(shù)量
• Expanding the dataset to include online behaviour — i.e. number of times ad was clicked on, IP location, time ad was uploaded, etc.
  擴(kuò)展數(shù)據(jù)集以包含在線行為，例如，廣告被點(diǎn)擊的次數(shù)，IP地址，廣告被上傳的時(shí)間等。

For all the code used to generate results, see my GitHub repository — https://github.com/shedoesdatascience/fraudanalytics

有關(guān)用于生成結(jié)果的所有代碼，請(qǐng)參見我的GitHub存儲(chǔ)庫— https://github.com/shedoesdatascience/fraudanalytics

翻譯自: https://towardsdatascience.com/identifying-fraudulent-job-advertisements-using-r-programming-230daa20aec7

欺詐行為識(shí)別

總結(jié)

以上是生活随笔為你收集整理的欺诈行为识别_使用R（编程）识别欺诈性的招聘广告的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。