By successfully solving the problem of forecasting, the processes in the work of various companies are optimized and savings are achieved. In this process, the analysis of time series data is of particular importance. Since the creation of Facebook's Prophet, and Amazon's DeepAR and CNN-QR forecasting models, algorithms have attracted a great deal of attention. The paper presents the application and comparison of the above algorithms for sales forecasting in distribution companies. A detailed comparison of the performance of algorithms over real data with different lengths of sales history was made. The results show that Prophet gives better results for items with a longer history and frequent sales, while Amazon's algorithms show superiority for items without a long history and items that are rarely sold. NTRODUCTION Successful sales forecasting mechanisms can have positive effects in many areas of business, and one of the basic aspects is stock optimization. In retail, wholesale and distribution companies, inventory optimization is one of the key aspects of business. Companies that maintain their stocks at an adequate and satisfactory level can save significant amounts of money, and at the same time their other processes, such as warehousing, commissioning, shipping, etc. are significantly improved. Stock optimization often does not have enough attention in a real environment. According to the detailed analysis presented by Bajrić [1], inventory management in the average company from Bosnia and Herzegovina is far from satisfactory. There are either too many products in the stock, so there is an unnecessary cost of keeping them, or not enough products, so there is a lost sales, cost of stopping production, replanting, switching to other products, breaking deadlines, returning to production of the original product and related costs. According to the mentioned research, stocks in the average Bosnian company can be reduced by an average of 25%.

Graph convolutional network (GCN) based approaches have achieved significant progress for solving complex, graph-structured problems. GCNs incorporate the graph structure information and the node (or edge) features through message passing and computes 'deep' node representations. Despite significant progress in the field, designing GCN architectures for heterogeneous graphs still remains an open challenge. Due to the schema of a heterogeneous graph, useful information may reside multiple hops away. A key question is how to perform message passing to incorporate information of neighbors multiple hops away while avoiding the well-known over-smoothing problem in GCNs. To address this question, we propose our GCN framework 'Deep Heterogeneous Graph Convolutional Network (DHGCN)', which takes advantage of the schema of a heterogeneous graph and uses a hierarchical approach to effectively utilize information many hops away. It first computes representations of the target nodes based on their 'schema-derived ego-network' (SEN). It then links the nodes of the same type with various pre-defined metapaths and performs message passing along these links to compute final node representations. Our design choices naturally capture the way a heterogeneous graph is generated from the schema. The experimental results on real and synthetic datasets corroborate the design choice and illustrate the performance gains relative to competing alternatives.

Traffic prediction is the cornerstone of an intelligent transportation system. Accurate traffic forecasting is essential for the applications of smart cities, i.e., intelligent traffic management and urban planning. Although various methods are proposed for spatio-temporal modeling, they ignore the dynamic characteristics of correlations among locations on road networks. Meanwhile, most Recurrent Neural Network (RNN) based works are not efficient enough due to their recurrent operations. Additionally, there is a severe lack of fair comparison among different methods on the same datasets. To address the above challenges, in this paper, we propose a novel traffic prediction framework, named Dynamic Graph Convolutional Recurrent Network (DGCRN). In DGCRN, hyper-networks are designed to leverage and extract dynamic characteristics from node attributes, while the parameters of dynamic filters are generated at each time step. We filter the node embeddings and then use them to generate a dynamic graph, which is integrated with a pre-defined static graph. As far as we know, we are the first to employ a generation method to model fine topology of dynamic graph at each time step. Further, to enhance efficiency and performance, we employ a training strategy for DGCRN by restricting the iteration number of decoder during forward and backward propagation. Finally, a reproducible standardized benchmark and a brand new representative traffic dataset are opened for fair comparison and further research. Extensive experiments on three datasets demonstrate that our model outperforms 15 baselines consistently.

In classification tasks, one may encounter a situation where the target class label is not equally distributed. Such a dataset can be termed Imbalanced data. Imbalance in data can be a blocker to train a data science model. In case of imbalance class problems, the model is trained mainly on the majority class and the model becomes biased towards the majority class prediction. Hence handling of imbalance class is essential before proceeding to the modeling pipeline.

In this article, we will discuss data cleaning and some insights from the analysis of fund investment companies. The investment company wants to know the top nine countries for investment from the data. The data is available here. We observe that the two data have column permalink and company permalink, in which the organization name at first index of both data frame is same but one is in capital and one is in lower case. Now, we will check the number of unique companies in both the data set.

Interviews are the most challenging part of getting any job especially for Data Scientist and Machine Learning Engineer roles where you are tested on Machine Learning and Deep Learning concepts. So, Given below is a short quiz that consists of 25 Questions consisting of MCQs(One or more correct), True-False, and Integer Type Questions to check your knowledge. Explanation: The derivative of Leaky RELU activation function h(z) is 1 only for z 0, while for z 0, it has a very small value. Explanation: Residuals are vertical offset and the sum of residuals is always zero. Explanation: For deciding class w1, the conditional Risk for w1 is smaller than w2.

Learning individualized treatment rules (ITRs) is an important topic in precision medicine. Current literature mainly focuses on deriving ITRs from a single source population. We consider the observational data setting when the source population differs from a target population of interest. We assume subject covariates are available from both populations, but treatment and outcome data are only available from the source population. Although adjusting for differences between source and target populations can potentially lead to an improved ITR for the target population, it can substantially increase the variability in ITR estimation. To address this dilemma, we develop a weighting framework that aims to tailor an ITR for a given target population and protect against high variability due to superfluous covariate shift adjustments. Our method seeks covariate balance over a nonparametric function class characterized by a reproducing kernel Hilbert space and can improve many ITR learning methods that rely on weights. We show that the proposed method encompasses importance weights and the so-called overlap weights as two extreme cases, allowing for a better bias-variance trade-off in between. Numerical examples demonstrate that the use of our weighting method can greatly improve ITR estimation for the target population compared with other weighting methods.

Current theoretical results on optimization trajectories of neural networks trained by gradient descent typically have the form of rigorous but potentially loose bounds on the loss values. In the present work we take a different approach and show that the learning trajectory can be characterized by an explicit asymptotic at large training times. Specifically, the leading term in the asymptotic expansion of the loss behaves as a power law $L(t) \sim t^{-\xi}$ with exponent $\xi$ expressed only through the data dimension, the smoothness of the activation function, and the class of function being approximated. Our results are based on spectral analysis of the integral operator representing the linearized evolution of a large network trained on the expected loss. Importantly, the techniques we employ do not require specific form of a data distribution, for example Gaussian, thus making our findings sufficiently universal.

Randomized clinical trials (RCTs) eliminate confounding but impose strict exclusion criteria that prevent sampling of the entire clinical population. Observational datasets are more inclusive but suffer from confounding. Difference in Differences (DD) eliminates confounding from observational data by comparing outcomes before and after treatment administration. However, the algorithm requires a parallel slopes assumption that may not apply in practice when confounding shifts across time. In this paper, we propose Synthesized Difference in Differences (SDD) that infers the correct (possibly non-parallel) slopes by linearly adjusting a conditional version of DD using additional RCT data. The algorithm achieves state of the art performance across multiple synthetic and real datasets even when the RCT excludes the majority of patients.

High-dimensional sparse generalized linear models (GLMs) have emerged in the setting that the number of samples and the dimension of variables are large, and even the dimension of variables grows faster than the number of samples. False discovery rate (FDR) control aims to identify some small number of statistically significantly nonzero results after getting the sparse penalized estimation of GLMs. Using the CLIME method for precision matrix estimations, we construct the debiased-Lasso estimator and prove the asymptotical normality by minimax-rate oracle inequalities for sparse GLMs. In practice, it is often needed to accurately judge each regression coefficient's positivity and negativity, which determines whether the predictor variable is positively or negatively related to the response variable conditionally on the rest variables. Using the debiased estimator, we establish multiple testing procedures. Under mild conditions, we show that the proposed debiased statistics can asymptotically control the directional (sign) FDR and directional false discovery variables at a pre-specified significance level. Moreover, it can be shown that our multiple testing procedure can approximately achieve a statistical power of 1. We also extend our methods to the two-sample problems and propose the two-sample test statistics. Under suitable conditions, we can asymptotically achieve directional FDR control and directional FDV control at the specified significance level for two-sample problems. Some numerical simulations have successfully verified the FDR control effects of our proposed testing procedures, which sometimes outperforms the classical knockoff method.