The transactions of goods and services between enterprise service providers are often driven by contracts and purchase orders. Every month thousands of invoices are billed to customers who settle them based on the usage of services. Considering the vast number of purchase orders that are signed, it requires considerable manual effort by the service provider to process and manage them. Moreover, the invoice’s billed data may not be maintained in the same cloud system as the purchase orders. This leads to complexity with data mapping between the two data sets. Sometimes the invoices may get into a dispute due to over exhaustion of allocated funds or may be billed to an expired purchase order. Hence managing the billing service is a huge undertaking along with increased cost.To address these challenges, we developed an order manage- ment system that transforms the monitoring of purchase orders to increase renewals as well as decrease disputes. The system includes an automated purchase order-invoice data mapping model along with a risk analytics model that evaluates the orders against the invoices billed. The output is the set of actionable and non actionable insights based on customer portfolio, risk level as well as market trends in usage of services. We illustrate our method with some promising results on data of one of the world’s largest IT service providers.

When Information Technology (IT) service providers cater services to their customers, there’s a common practice to document the intent of buyer to purchase them through a Purchase Order (PO). The details of a PO document are complex in nature, as they include hierarchical structure of sub-services, price points over the duration of PO, base setup cost, billing frequency, renewal terms etc. In large enterprises with high volume PO’s, the traditional approach of managing PO’s with their invoices involved a great deal of inefficiency and labor-intensive manual work due to lack of automation and disconnected processing. As a result, they often suffered tedious manual monitoring and failed PO renewals, resulting in delays and added costs.In this paper, we propose a joint time-series learning framework that tackles the high dimensionality PO data. It identifies the metrics to monitor, while using time-series prediction coupled with tone analysis to reach out to customers, in order to maximise PO renewals. We demonstrate utility of our approach by implementing our method on a dataset from a global IT service provider that indicates encouraging results.

Sellers and executives who maintain a bidding pipeline of sales engagements with multiple clients for many opportunities significantly benefit from data-driven insight into the health of each of their bids. There are many predictive models that offer likelihood insights and win prediction modeling for these opportunities. Currently, these win prediction models are in the form of binary classification and only make a prediction for the likelihood of a win or loss. The binary formulation is unable to offer any insight as to why a particular deal might be predicted as a loss. This paper offers a multi-class classification model to predict win probability, with the three loss classes offering specific reasons as to why a loss is predicted, including no bid, customer did not pursue, and lost to competition. These classes offer an indicator of how that opportunity might be handled given the nature of the prediction. Besides offering baseline results on the multi-class classification, this paper also offers results on the model after class imbalance handling, with the results achieving a high accuracy of 85% and an average AUC score of 0.94.

In the face of emergent external factors (e.g., supply chain disruptions or public health crises like COVID-19), businesses must adapt their business model quickly in order to ensure service continuity. However, providing recommendations regarding changes should be made to the business model is a challenging problem. First, it requires details of interactions between different components of the business (e.g., service offerings, inventory, staffing, demand) to understand what possible courses of action will have the most business impact. Second, automated models may provide recommendations on changes required in the business operations. However, with lack of human insight, it will be hard to verify the feasibility of these recommendations. Third, a generic model may not be able to provide good recommendations for diverse set of business models. Fourth, the model may not have enough features or training data to provide good recommendations.In this paper, we propose a novel approach to provide actionable items that can be recommended to business users given their business features and recommendations given to businesses in similar domain. Here we first use clustering to find the business domain and similar feature set of the domain. Then, we build a machine-learning model with explainable insights to provide recommendations on different business actions that can be taken to ensure business operations in the face of emergent external factors. Next we augment our approach with human-in-the-loop to improve its performance. Finally, we federate the machine-learning model in a similar domain to add more explainable and trusted insights and recommendations by other businesses. We describe our method, illustrate its utility with results from our implementation, and discuss areas for future work.

Although widely used in various areas, the Deep Neural Network suffers from the lack of interpretability. Existing works usually focus on one data instance and the found explanations are thus limited. We argue that a good understanding of a model should contain both systematic explanations of model behavior and effective detection of its vulnerability. Particularly we propose to use backward deduction to achieve these two goals. Given a constraint on the model output, the deduction backtracks the architecture to find corresponding data ranges in the input. In each layer, depending on the type, specific rules and/or algorithms are developed. The resulted ranges in the end can be interpreted by sampling exemplary data from them. In experiment we show that with different strategies in selecting the data ranges, the sampled fake data can either explain the model or reflect its vulnerability.

Emotion analysis of online customer service conservation is important for good user experience and customer satisfaction. However, conventional metrics do not fit this application scenario. In this work, by collecting and labeling online conversations of customer service on Twitter, we identify 8 new metrics, named as tones, to describe emotional information. To better interpret each tone, we extend the Latent Dirichlet Allocation (LDA) model to Tone LDA (T-LDA). In T-LDA, each latent topic is explicitly associated with one of three semantic categories, i.e., tone-related, domain-specific and auxiliary. By integrating tone label into learning, T-LDA can interfere the original unsupervised training process and thus is able to identify representative tone-related words. In evaluation, T-LDA shows better performance than baselines in predicting tone intensity. Also, a case study is conducted to analyze each tone via T-LDA output.

Efficiency of large-scale learning is a hot topic in both academic and industry. The stochastic gradient descent (SGD) algorithm, and its extension mini-batch SGD, allow the model to be updated without scanning the whole data set. However, the use of approximate gradient leads to the uncertainty issue, slowing down the decreasing of objective function. Furthermore, such uncertainty may result in a high frequency of meaningless update on the model, causing a communication issue in parallel learning environment. In this work, we develop a batch-adaptive stochastic gradient descent (BA-SGD) algorithm, which can dynamically choose a proper batch size as learning proceeds. Particularly on the basis of Taylor extension and central limit theorem, it models the decrease of objective value as a Gaussian random walk game with rebound. In this game, a heuristic strategy of determining batch size is adopted to maximize the utility of each incremental sampling. By evaluation on multiple real data sets, we demonstrate that by smartly choosing the batch size, the BA-SGD not only conserves the fast convergence of SGD algorithm but also avoids too frequent model updates.