A Blended Deep Learning Approach for Predicting User Intended Actions

Abstract

User intended actions are widely seen in many areas. Forecasting these actions and taking proactive measures to optimize business outcome is a crucial step towards sustaining the steady business growth. In this work, we focus on predicting attrition, which is one of typical user intended actions. Conventional attrition predictive modeling strategies suffer a few inherent drawbacks. To overcome these limitations, we propose a novel end-to-end learning scheme to keep track of the evolution of attrition patterns for the predictive modeling. It integrates user activity logs, dynamic and static user profiles based on multi-path learning. It exploits historical user records by establishing a decaying multi-snapshot technique. And finally it employs the precedent user intentions via guiding them to the subsequent learning procedure. As a result, it addresses all disadvantages of conventional methods. We evaluate our methodology on two public data repositories and one private user usage dataset provided by Adobe Creative Cloud. The extensive experiments demonstrate that it can offer the appealing performance in comparison with several existing approaches as rated by different popular metrics. Furthermore, we introduce an advanced interpretation and visualization strategy to effectively characterize the periodicity of user activity logs. It can help to pinpoint important factors that are critical to user attrition and retention and thus suggests actionable improvement targets for business practice. Our work will provide useful insights into the prediction and elucidation of other user intended actions as well.

An Interpretable Pipeline for Identifying At-Risk Students

“…To demonstrate the effectiveness of the proposed model, in this section, we introduce several machine learning models that are commonly used as baseline models in the literature. Random Forest (RF) has been demonstrated with a good performance in dropout prediction and is frequently deployed in the industry (Tan et al., 2018). During the prediction process, multiple decision trees are constructed on random attributes to produce generalizable outcomes, which greatly avoids the potentials of overfitting during the training process (Jayaprakash et al., 2020).…”

“…Recently, increasing attention has been paid to the applications of deep learning approaches in analyzing students' online learning performance due to the capabilities of dealing with sequential data. Specifically, the approaches such as Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) Network (Fei & Yeung, 2015;Tan et al, 2018;Xiong et al, 2019) have been applied to explore the sequential correlations between students' learning behaviors and build models to predict the learning behaviors that are more likely to be conducted in future and students' learning performance. Specifically, Tan et al (2018) proposed a multi-path learning-based scheme to track students' learning behaviors for predicting the probabilities of attrition.…”

“…Specifically, the approaches such as Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) Network (Fei & Yeung, 2015;Tan et al, 2018;Xiong et al, 2019) have been applied to explore the sequential correlations between students' learning behaviors and build models to predict the learning behaviors that are more likely to be conducted in future and students' learning performance. Specifically, Tan et al (2018) proposed a multi-path learning-based scheme to track students' learning behaviors for predicting the probabilities of attrition. This approach employed a neural network structure to keep track of the trends of students' attrition.…”

“…Generally, this is caused by inherent mechanisms while building the models: On the one hand, most machine learning models (i.e., Logistic Regression, Decision Tree) used currently were built without considering the associations among the learning behaviors. Moreover, the feature extraction approaches (Liang et al., 2016) that are commonly used by the machine learning models are not built for dealing with temporal-related information, which also impacts the capabilities of these models in capturing the time-sensitive information in the dataset; On the other hand, while the approaches (Fei & Yeung, 2015; Tan et al., 2018) such as Recurrent Neural Networks, Long-Short Term Memory Networks are built for dealing with the sequential information, they always have to maintain complex network structures to model the associations among the high dimensional data, which may result in high computational complexity and low output interpretability.…”

An Enriched Time-Series Forecasting Framework for Long-Short Portfolio Strategy

“…More hybrid models for stock market forecasting can be found in [31], [33], [35]- [37]. Recently, the hybridization of different kinds of neural networks, which can accommodate heterogeneous input features, has been applied to other areas successfully, e.g., success prediction on crowdfunding [38] and user intended action prediction [39]. These promising results motivate us to blend different kinds of neural networks for characterizing the heterogeneous input features to predict stock returns.…”