This full research paper presents a systematic analysis of 10 years' student performance data of Computer Science (CS) majors at San Francisco State University, a public 4-year degree-granting university, aiming to address the ongoing challenges of early dropouts and low graduation rate. The main objective is two-fold: (1) gain a comprehensive understanding of how the existing curriculum has been supporting (or hindering) students' progress towards graduation; and (2) suggest data-informed curricular changes. To this end, we utilize both explorative statistical analysis and data mining/machine learning approaches to first learn how individual courses and the prescribed course sequences influence a student's dropout/graduation status, and then build machine learning models to interpret/validate the observed interdependency among key courses in the current curriculum. Such patterns/models are consequently utilized to suggest impactful curricular changes towards reducing early dropouts and improving the overall student success as measured by graduation with a CS degree. One main finding of this research is that a successful CS student needs to excel in both critical thinking and core CS skills. To help students gain critical thinking skills, it is essential to strengthen the presence of mathematics and physics courses in the CS curriculum. Furthermore, our results suggest that CS students without a solid math foundation before starting their college career should complete a remedial math course earlier than putting it off for later. Moreover, before students advance to the second half of their CS study to gain core CS knowledge/skills (e.g., operating systems), they should complete the required physics class. Finally, we observe that it is necessary to introduce new prerequisite requirements among upper-level CS courses, for example, Operating Systems as a prerequisite to an upper-level CS core course on programming theories.
The Anatomical Therapeutic Chemical (ATC) Code System is a World Health Organization (WHO) proposed classification that assigns codes to compounds based on their therapeutic, pharmacological and chemical characteristics as well as the in-vivo site of activity. The ability to predict the ATC code of an arbitrary compound with high accuracy can go a long way in selecting molecules for lead identification. We propose a computational approach to this problem that utilizes a natural pharmacological constraint, namely, that anatomical-therapeutic biological activity of certain types must preclude activities of many other types. The method proposed here utilizes machine learning in a tiered architecture; prediction of the ATC code at a certain level is constrained by the ATC code at the higher levels. Using this learning architecture, we have built classifiers that incorporate information from a compound's structure, as well as its chemical and protein interactions. The proposed approach has been validated using 2335 drugs from the ChEMBL database in both cross-validation and test setting. The prediction accuracy obtained with this approach is 78.72% and is comparable or better than the prediction accuracy of other methods at the state of the art.
The low success rate and high cost of drug discovery requires the development of new paradigms to identify molecules of therapeutic value. The Anatomical Therapeutic Chemical (ATC) Code System is a World Health Organization (WHO) proposed classification that assigns multi-level codes to compounds based on their therapeutic, pharmacological and chemical characteristics as well as the in-vivo sites(s) of activity. The ability to predict ATC codes of compounds can assist in creation of high-quality chemical libraries for drug screening and in applications such as drug repositioning. We propose a machine learning architecture called tiered learning for prediction of ATC codes that relies on the prediction results of the higher levels of the ATC code to simplify the predictions of the lower levels.The proposed approach was validated using a number of compounds in both cross-validation and test setting. The validation experiments compared chemical descriptors, initialization methods and classification algorithms. The prediction accuracy obtained with tiered learning was found to be either comparable or better than that of established methods. Additionally, the experiments demonstrated the generalizability of the tiered learning architecture, in that its use was found to improve prediction rates for a majority of machine learning algorithms when compared to their stand-alone application.The basis of our approach lies in the observation that anatomical-therapeutic biological activity of certain types typically precludes activities of many other types. Thus, there exists a characteristic distribution of the ATC codes, which can be leveraged to limit the search-space of possible codes that can be ascribed at a particular level once the codes at the preceding levels are known. Tiered learning utilizes this observation to constrain the learning space for ATC codes at a particular level based on the ATC code at higher levels. This simplifies the prediction and allows for improved accuracy.
'%3e%3cpath fill='%23012169' d='M0 0v30h60V0z'/%3e%3cpath stroke='%23fff' stroke-width='6' d='m0 0 60 30m0-30L0 30'/%3e%3cpath stroke='%23C8102E' stroke-width='4' d='m0 0 60 30m0-30L0 30' clip-path='url(%23b)'/%3e%3cpath stroke='%23fff' stroke-width='10' d='M30 0v30M0 15h60'/%3e%3cpath stroke='%23C8102E' stroke-width='6' d='M30 0v30M0 15h60'/%3e%3c/g%3e%3c/svg%3e)
