Agile and Accurate CTR Prediction Model Training for Massive-Scale Online Advertising Systems

With the development of E-commerce, online advertising began to thrive and has gradually developed into a new mode of business, of which Click-Through Rates (CTR) prediction is the essential driving technology. Given a user, commodities and scenarios, the CTR model can predict the user’s click probability of an online advertisement. Recently, great progress has been made with the introduction of Deep Neural Networks (DNN) into CTR. In order to further advance the DNN-based CTR prediction models, this paper introduces a new model of FO-FTRL-DCN, based on the prestigious model of Deep&Cross Network (DCN) augmented with the latest optimization technique of Follow The Regularized Leader (FTRL) for DNN. The extensive comparative experiments on the iPinYou datasets show that the proposed model has outperformed other state-of-the-art baselines, with better generalization across different datasets in the benchmark.

Download Full-text

Improved protein contact prediction using dimensional hybrid residual networks and singularity enhanced loss function

10.1101/2021.05.10.443415 ◽

2021 ◽

Author(s):

Yunda Si ◽

Chengfei Yan

Keyword(s):

Prediction Model ◽

Loss Function ◽

Prediction Models ◽

Model Performance ◽

Great Success ◽

Contact Prediction ◽

Reference Models ◽

Residual Learning ◽

Model Training ◽

Residual Block

Deep residual learning has shown great success in protein contact prediction. In this study, a new deep residual learning-based protein contact prediction model was developed. Comparing with previous models, a new type of residual block hybridizing 1D and 2D convolutions was designed to increase the effective receptive field of the residual network, and a new loss function emphasizing the easily misclassified residue pairs was proposed to enhance the model training. The developed protein contact prediction model referred to as DRN-1D2D was first evaluated on 105 CASP 11 targets, 76 CAMEO hard targets and 398 membrane proteins together with two in house-developed reference models based on either the standard 2D residual block or the traditional BCE loss function, from which we confirmed that that the dimensional hybrid residual block and the singularity enhanced loss function can both be employed to improve the model performance for protein contact prediction. DRN-1D2D was further evaluated on 39 CASP 13 and CASP 14 free modeling targets together with the two reference models and four state-of-the-art protein contact prediction models including DeepCov, DeepCon, RaptorX-Contact and TripleRes. The result shows that DRN-1D2D consistently achieved the best performance among all these models.

Download Full-text

A Prediction Model for Regional Carbon Emissions Based on GRU Networks

10.21203/rs.3.rs-767805/v1 ◽

2021 ◽

Author(s):

Jun Meng ◽

Gangyi Ding ◽

Laiyang Liu ◽

Zheng Guan

Keyword(s):

Prediction Model ◽

Carbon Emissions ◽

Missing Values ◽

Environmental Data ◽

Sensor Data ◽

Data Set ◽

Bp Network ◽

Gross Error ◽

Model Training ◽

Exponential Smoothing Method

Abstract In this study, a data-driven regional carbon emissions prediction model is proposed. The Grubbs criterion is used to eliminate the gross error data in carbon emissions sensor data. Then, according to the nearby valid data, the exponential smoothing method is used to interpolate the missing values to generate the continuous sequence for model training. Finally, the GRU network, which is a deep learning method, is used to process these sequential standardized data to obtain the prediction model. In this paper, the wireless carbon sensor network monitoring data set from August 2012 to April 2014 trained and evaluated the prediction model, and compared with the prediction model based on BP network. The experimental results prove the feasibility of the research method and related technical approaches, and the accuracy of the prediction model, which provides a method basis for the nowcasting of carbon emissions and other greenhouse gas environmental data.

Download Full-text

Company employee quality evaluation model based on BP neural network

Journal of Intelligent & Fuzzy Systems ◽

10.3233/jifs-189428 ◽

2020 ◽

pp. 1-10

Author(s):

Tsui-Yuan Tseng ◽

Qinglan Luo

Keyword(s):

Neural Network ◽

Prediction Model ◽

Relative Error ◽

Bp Neural Network ◽

Quality Evaluation ◽

Evaluation Model ◽

Performance Data ◽

Average Relative Error ◽

Model Training ◽

Quality Evaluation Model

With the development of science and technology and the continuous improvement of people’s living standards, the traditional staff quality evaluation can no longer meet the needs of production and life, and the BP neural network has also appeared many shortcomings in practical applications. This article mainly studies the company’s employee quality evaluation model based on BP neural network. This article first collects and preprocesses employees’ usual performance data, and then predicts their corresponding quality scores based on BP neural network. And use MATLAB software to simulate the constructed prediction model, and finally develop a complete set of employee performance data prediction system based on this model, so as to achieve the purpose of employee quality evaluation. The experimental data in this paper shows that the average relative error of model training output tends to be stable. After the 40th iteration of training, the average relative error of model training can reach 0.0128. After the prediction model training was completed, 15 sets of verification samples were used to verify the model. The verification results found that the average relative error of the model converged, so the model did not overfit. Experimental results show that although BP neural network has two excellent functions of adaptive and nonlinear approximation, it can solve the complex nonlinear relationship between normal performance and overall performance. But BP neural network still has its own inevitable shortcomings in some aspects. For example the redundancy between the employee scoring sample data; the problem that the input variable dimensionality is too high, which leads to the low efficiency of the model; the fuzzy neural network is easy to fall into the local optimum and it is difficult to find the global optimum.

Download Full-text

Analysis of College Student Registration Management and Change Prediction Based on Mutated Fuzzy Neural Network Algorithm

Computational Intelligence and Neuroscience ◽

10.1155/2021/7097425 ◽

2021 ◽

Vol 2021 ◽

pp. 1-9

Author(s):

Yao Wang ◽

Lie Jiao ◽

Chunzhi Liu

Keyword(s):

Neural Network ◽

Prediction Model ◽

Fuzzy Neural Network ◽

Principal Component ◽

Correct Prediction ◽

Proposed Model ◽

School Teaching ◽

Fuzzy Neural ◽

Optimization Function ◽

Model Training

Nowadays, a large number of students' academic registrations change every year in universities, but most of these cases are recorded and mathematically and statistically analysed through forms or systems, which are cumbersome and difficult to find some potential information in them. Therefore, timely and effective prediction of student registration changes and early warning of student registration changes by technical means is an important part of university registration management. At present, relevant research is mostly based on mathematical statistical analysis methods such as students' current credit evaluation or course score averages and seldom uses data mining and other technical methods for in-depth research. In this paper, we propose a mutated fuzzy neural network (MFNN) based prediction model for student registration changes in colleges and universities, which can provide supplementary reference decisions for school registration management for school teaching managers. In this paper, we first construct the corresponding prediction model of academic registration variation, define the relevant parameters, and model the optimization problem and propose the objective optimization function. Second, the proposed model is optimized by adding principal component analysis (PCA) to the original model to improve the efficiency of model training and the correct prediction rate. It is verified that the proposed model can effectively predict individual students' academic registration changes with a prediction accuracy of nearly 92.91%.

Download Full-text